Intelligent Manufacturing and Energy Sustainability: Proceedings of ICIMES 2022 9811984964, 9789811984969

Citation preview

Smart Innovation, Systems and Technologies 334

Amaranadha Reddy Manchuri Deepak Marla V. Vasudeva Rao   Editors

Intelligent Manufacturing and Energy Sustainability Proceedings of ICIMES 2022

Smart Innovation, Systems and Technologies Volume 334

Series Editors Robert J. Howlett, Bournemouth University and KES International, Shoreham-by-Sea, UK Lakhmi C. Jain, KES International, Shoreham-by-Sea, UK

The Smart Innovation, Systems and Technologies book series encompasses the topics of knowledge, intelligence, innovation and sustainability. The aim of the series is to make available a platform for the publication of books on all aspects of single and multi-disciplinary research on these themes in order to make the latest results available in a readily-accessible form. Volumes on interdisciplinary research combining two or more of these areas is particularly sought. The series covers systems and paradigms that employ knowledge and intelligence in a broad sense. Its scope is systems having embedded knowledge and intelligence, which may be applied to the solution of world problems in industry, the environment and the community. It also focusses on the knowledge-transfer methodologies and innovation strategies employed to make this happen effectively. The combination of intelligent systems tools and a broad range of applications introduces a need for a synergy of disciplines from science, technology, business and the humanities. The series will include conference proceedings, edited collections, monographs, handbooks, reference books, and other relevant types of book in areas of science and technology where smart systems and technologies can offer innovative solutions. High quality content is an essential feature for all book proposals accepted for the series. It is expected that editors of all accepted volumes will ensure that contributions are subjected to an appropriate level of reviewing process and adhere to KES quality principles. Indexed by SCOPUS, EI Compendex, INSPEC, WTI Frankfurt eG, zbMATH, Japanese Science and Technology Agency (JST), SCImago, DBLP. All books published in the series are submitted for consideration in Web of Science.

Amaranadha Reddy Manchuri · Deepak Marla · V. Vasudeva Rao Editors

Intelligent Manufacturing and Energy Sustainability Proceedings of ICIMES 2022

Editors Amaranadha Reddy Manchuri Department of Mechanical Engineering Malla Reddy College of Engineering and Technology Hyderabad, India

Deepak Marla Department of Mechanical Engineering Indian Institute of Technology Bombay Mumbai, Maharashtra, India

V. Vasudeva Rao Department of Mechanical and Industrial Engineering, College of Science, Engineering and Technology University of South Africa Pretoria, South Africa

ISSN 2190-3018 ISSN 2190-3026 (electronic) Smart Innovation, Systems and Technologies ISBN 978-981-19-8496-9 ISBN 978-981-19-8497-6 (eBook) https://doi.org/10.1007/978-981-19-8497-6 © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore

ICIMES-2022 Committees

Conference Committee Chief Patron Sri. CH. Malla Reddy, Founder Chairman, MRGI

Patrons Sri. CH. Mahendar Reddy, Secretary, MRGI Sri. CH. Bhadra Reddy, President, MRGI

Conference Chair Dr. V. S. K. Reddy, Principal

Honorary Chair Dr. Lakshmi C. Jain, University of Sydney, Sydney, Australia

Publication Chair Dr. Suresh Chandra Satapathy, Professor, KIIT, Bhubaneswar, India v

vi

ICIMES-2022 Committees

Convener Dr. S. Srinivasa Rao, Principal, MRCET

Co-convener Dr. P. H. V. Sesha Talpa Sai, Director, R&D, MRCET

Organizing Chair Dr. Amaranadha Reddy Manchuri, Professor & HOD, Mechanical Engineering

Organizing Secretary Dr. Srikar Potnuru, Associate Professor, Mechanical Engineering

Coordinator Prof. Vennam Gopala Krishan, Assistant Professor, Mechanical Engineering

Editorial Board Dr. Amaranadha Reddy Manchuri, Malla Reddy College of Engineering and Technology, India Dr. Deepak Marla, Indian Institute of Technology Bombay, India Dr. V. Vasudeva Rao, University of South Africa

International Advisory Committee Dr. Lakshmi C. Jain, University of Sydney, Sydney, Australia Dr. Narayanan Kulathuramaiyer, Universiti Malaysia Sarawak, Malaysia Dr. Abu Saleh Ahmed, Universiti Malaysia Sarawak, Malaysia

ICIMES-2022 Committees

vii

Dr. Shahrol Mohamaddan, Shibaura Institute of Technology, Japan Dr. Sinin Hamdan, Universiti Malaysia Sarawak, Malaysia Dr. Jaesool Shim, Yeunagnam University, South Korea Dr. Amiya Bhaumik, Lincoln University College, Malaysia Dr. Bhaskar Kura, University of New Orleans, LA, USA Dr. Devarayapalli K. C., Yeunagnam University, South Korea Dr. Raja V. Pulikollu, Electric Power Research Institute, North Carolina, USA Dr. Angel Sanz Anderes, UPM, Madrid, Spain Dr. S. V. Prabhakar, Yeunagnam University, South Korea Dr. Yequing Bao, University of Alabama, USA Dr. Sabastian Franchini, UPM, Madrid, Spain

National Advisory Committee Dr. G. Balu, DOAD, DRDL, Telangana, India Dr. Koteshwara Rao Kandula, Satish Dhawan Space Centre, Sriharikota (ISRO), India Dr. Susanta Kumar Sahoo, NIT Rourkela, Odisha, India Dr. Swayam Bikash Mishra, KIIT, Bhubaneswar, India Dr. Ajit Behera, NIT Rourkela, Odisha, India Dr. Jose Immanuel, IIT Bhilai, Chhattisgarh, India Dr. P. K. Jain, Centre for Carbon Materials, ARCI, Hyderabad, Telangana, India Dr. K. Vijay Kumar Reddy, JNTU Hyderabad, Telangana, India Dr. S. Surya Kumar, IIT Hyderabad, Telangana Dr. Deepak K. Ojha, IIT Roorkee, Uttarakhand Dr. K. S. Reddy, Indian Institute of Technology Madras, Tamil Nadu, India Dr. K. Uday Kumar, NIT Warangal, Telangana Dr. G. Raghavendra, National Institute of Technology Warangal, Telangana, India Dr. Chinmay, NIT Silchar, Assam, India Dr. D. Haranath, NIT Warangal, Telangana Dr. Swami Naidu, National Institute of Technology Raipur, Chhattisgarh, India Dr. Rajesh Kumar Rajaboina, NIT Warangal, Telangana, India Dr. Jaya Chandra Bingi, IITDM, Kancheepuram, India Dr. P. Narsimha Reddy, SNIST, Hyderabad, Telangana, India Dr. U. S. Paul Russel, Air India, India

Industry Advisory Committee Dr. Sailesh K. Kharade, Cyient, India Mr. Sunil Maheshwari, Sr. Regional Manager, Adroitec Engineering Solutions Pvt. Ltd., India

viii

ICIMES-2022 Committees

Mr. Bhusan Prasad M, Vice President, Renewable Energy Systems Ltd., India Mr. Uddagiri Vidyasagar, TCS, Hyderabad, India Mr. Uma Shankar, Farm Division, Mahindra & Mahindra, Zaheerabad, India Mr. Sandeep Arora, Associate Director, Verizon, India Mr. Ramana Rao P. V., ADM, SMC Corporation India Pvt. Ltd. M/s. Ramky Enviro Engineers Ltd.

Institute Committee Prof. (Dr.) V. Madhusudhana Reddy Prof. (Dr.) T. Siva Kumar Prof. (Dr.) T. Lokeswara Rao Prof. (Dr.) D. Damodar Reddy Prof. (Mr.) Y. Dilip Kumar Prof. (Ms.) S. Deepthi Prof. (Ms.) K. Akhila Prof. (Mr.) M. Harish

Preface

The International Conference on Intelligent Manufacturing and Energy Sustainability (ICIMES-2022) was successfully organized by Malla Reddy College of Engineering and Technology, an UGC Autonomous Institution, during June 24–25, 2022, at Hyderabad. The objective of this conference was to provide opportunities for the researchers, academicians and industry persons to interact and exchange the ideas, experience and gain expertise in the cutting-edge technologies pertaining to Industry 5.0. Research papers were received and subjected to a rigorous peer review process with the help of editorial board, program committee and external reviewers. The editorial committee has finally accepted 21% manuscripts for publication in a single volume with Springer SIST series. Our sincere thanks to guest of honor and keynote session by Prof. Nishi Ryuhei— First Secretary (Science and Technology), Japan Embassy. Our outmost thanks to all our session chairs for their immense support. 1. Dr. D. K. Charyulu, Research Associate, School of Nano and Materials Science and Engineering, Kyungpook National University, Sangju, Gyeongbuk, South Korea. 2. Dr. K. Uday Kumar, Assistant Professor, Department of Physics, NIT Waranagal. 3. Dr. G. Bala Narasimha, Assistant Professor, Malla Reddy University, Hyderabad 4. Dr. J. Trinath, Assistant Professor, Malla Reddy University, Hyderabad. 5. Dr. Srikar Potnuru, Assistant Professor, Department of Mechanical Engineering, Malla Reddy College of Engineering and Technology, Hyderabad. 6. Dr. T. Siva Kumar, Associate Professor, Department of Mechanical Engineering, Malla Reddy College of Engineering and Technology, Hyderabad. 7. Dr. B. Sandhya Rani, Assistant Professor, Department of Mechanical Engineering, Malla Reddy College of Engineering and Technology, Hyderabad. 8. Dr. D. Damodara Reddy, Assistant Professor, Department of Mechanical Engineering, Malla Reddy College of Engineering and Technology, Hyderabad. We are indebted to the editorial board, program committee and external reviewers who have produced critical reviews in a short time. We would like to express our special gratitude to publication chair, Dr. Suresh Chandra Satapathy, Professor, ix

x

Preface

KIIT, Bhubaneswar, for his valuable support and encouragement till the successful conclusion of the conference. We express our heartfelt thanks to our chief patron, Sri. CH. Malla Reddy, founder chairman, MRGI; patrons Sri. CH. Mahendar Reddy, secretary, MRGI; Sri. CH. Bhadra Reddy, president, MRGI; conference chair, Dr. V. S. K. Reddy; convener, Dr. S. Srinivasa Rao; organizing chair, professor Dr. Amaranadha Reddy Manchuri; organizing secretary, Dr. Srikar Potnuru; and coordinator, Mr. V. Gopala Krishna, for their valuable contribution to successful conduct the conference. Last but certain1ly not least, our special thanks to all the authors without whom the conference would not have taken place. Their technical contributions have made our proceedings rich and praiseworthy. Hyderabad, India Mumbai, India Pretoria, South Africa

Dr. Amaranadha Reddy Manchuri Dr. Deepak Marla Dr. V. Vasudeva Rao

Contents

1

2

A Hybrid Optimal Technique for Road Extraction Using Entropy Rate Super-Pixel Segmentation and Probabilistic Neural Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. Subhashini and V. B. S. Srilatha Indira Dutt Meta Heuristic-Based Community Detection of Social Network Using Cuckoo with InfoMap Algorithm . . . . . . . . . . . . . . . . . S. Devi, M. Rajalakshmi, S. Saranya, and J. Shana

3

Smart Mirror Using Raspberry Pi 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . Desu Sai Pranav Reddy, Y. Sreevatsal Pranav, Padmavathi Kora, and V. Arvind

4

Garbage Bin Alert System Using Arduino UNO for Smart Cities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Gangidi Harathi, Gudise Anusha, G. Vinay Raj, and Sowjanya Ramisetty

5

LabVIEW-Based Temperature Control Using Fuzzy Logic Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. H. Sai Krishna, S. Srinivasulu Raju, B. N. V. S. S. R. Dhanush, K. Harsha Vardhan, and K. R. M. V. Ganesh

6

A Survey Towards Implementing Smart Campus . . . . . . . . . . . . . . . . . Anakhi Hazarika, K. D. K. Ajay, Nemani Subash, G. Srinivasa Yeshwanth, Lanka Raju, P. Kushal Swarup, S. K. Shireen Kasuar, and A. T. Antony

7

Design and Analysis of Stepped Impedance Feed Elliptical Patch Antenna . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . P. Sree Lakshmi, A. Deepak, Suresh Kumar Muthuvel, and Ch. Amarnatha Sarma

1

15 25

35

43

55

63

xi

xii

8

9

Contents

Wavelet-Based Colon Polyp Detection Using Support Vector Machine Classifier . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . B. Jyothi, M. Sucharitha, and Anitha Patibandla Probable Deviation Outlier-Based Classification of Obesity with Eating Habits and Physical Condition . . . . . . . . . . . . . . . . . . . . . . M. Shyamala Devi, P. S. Ramesh, Aparna Joshi, K. Maithili, and A. Prem Chand

10 Design and Implementation of Smart Home Automation System Using the Proteus Design Tool . . . . . . . . . . . . . . . . . . . . . . . . . . . L. Niranjan, Husna Tabassum, B. Sreekantha, T. Pushpa, and Mantri Gayatri

71

81

95

11 A Novel Approach to Prognosticate CKD Using a Supervised and Unsupervised Learning Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . 107 S. Ashwathi, A. Swamy Goud, L. Niranjan, B. Sreekantha, and J. Suneetha 12 Deep Network Architectures for Object Detection and Tracking: A Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 Chinthakindi Kiran Kumar, Gaurav Sethi, and Kirti Rawal 13 Smart Vision of IoT: Semantic Approach of Data Analysis and Data Analytics in Agriculture 4.0 . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 Smita Mane and Vaibhav E. Narawade 14 Hybrid Boost Converter Integrated Seven-Level MLI Fed PMSM Drive with Closed-Loop Speed Control . . . . . . . . . . . . . . . . . . . 139 J. A. Ganeswari and R. Kiranmayi 15 A Braille Learning Device for the Visually Impaired . . . . . . . . . . . . . . 151 Pranav Sakre, Sanika More, and Shruti Dodani 16 Handle the Sybil Attack Using Hash Technique in Vehicular Ad Hoc Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 A. Mummoorthy, N. S. Gowri Ganesh, R. Roopa Chandrika, and P. Swetha 17 Enhancing Recommender Systems Using Sentiment and Emotion Analysis of Reviews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167 Vamshi Krishna Dammoju, M. Samba Sivudu, M. Jayapal, and S. Shanthi 18 Prediction of Chronic Heart Disease using Machine Learning . . . . . 177 N. R. Rajalakshmi, J. Santhosh, J. Arun Pandian, and Mahmoud Alkhouli

Contents

xiii

19 Data Aegis Using Chebyshev Chaotic Map-Based Key Authentication Protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 Mohammed Abdul Lateef, C. Atheeq, Mohd Abdul Rahman, and Mohammed Abdul Faizan 20 MRAS for Induction Motor Using Fuzzy-PI Controller . . . . . . . . . . . 197 U. Arun Kumar, D. Maladhi, S. Gunasekaran, and S. Nandakumar 21 Anomaly Detection from Video Surveillances Using Adaptive Convolutional Neural Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205 Deepak Mane, Prashant Kumbharkar, Poonam Pawar, Karishma Katkar, Siddhali Shah, and Khushi Jamwal 22 Applications of Classification and Recommendation Techniques to Analyze Soil Data and Water Using IOT . . . . . . . . . . . 215 S. Thaiyalnayaki, Rakoth Kandan Sambandam, Thimmapuram Hima Sekhar Reddy, Alla Harsha Vardhan Reddy, Katla Naga Sai, and Vayakanti Srinivas 23 A Review on Epileptic Seizure Detection and Prediction . . . . . . . . . . 225 Suresh Nalla and Seetharam Khetavath 24 Sophisticated Machine Learning Algorithms for Pre-investigation of Heart Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . 233 Surfraz Mitegar, M. Samba Sivudu, and Giridhar Akula 25 Analysis Six Sigma Implementation in Power Plants . . . . . . . . . . . . . . 247 Telugu Sudhakar, B. Anjaneya Prasad, and K. Prahlada Rao 26 Blockchain: A Safe Way to Transfer Signatures in a Distributed Intrusion Detection System . . . . . . . . . . . . . . . . . . . . . . 261 Y. Madhusekhar, P. Sandhya Priyanka, Deena Babu Mandru, and T. Srikanth 27 Influence of MoS2 with TiC on the Tribological and Wear Properties of Hybrid Aluminum Composites . . . . . . . . . . . . . . . . . . . . . 275 Ramanan Gopalakrishnan, Bino Prince Raja Dennis, Neela Rajan Rajadurai Ramakrishnan, and Ajith Raj Rajendran 28 Optimization of Process Parameters in Friction Stir Welded AA6063 and AA7075 Joints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 285 Jalli Emmanuel, Prasada Raju Kantheti, G. Balakrishna, M. Vijay Daniel, and B. Durga Prasad 29 Experimental Investigation on Combustion and Emission Characteristics of Butanol Gasoline Fuelled SI Engine . . . . . . . . . . . . 295 S. Suryaprakash and S. Srihari

xiv

Contents

30 Preparation and Characterization of Zr-Nanofiller-Based PEO/PVDF-HFP-Mg Polymer Electrolyte Membranes for Battery Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307 N. Kundana, M. Venkatapathy, V. Neeraja, Chandra Sekhar Espenti, V. Madhusudhana Reddy, and A. Mallikarjun 31 Analysis of CO2 Refrigeration System in a Super Market: An Experimental Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317 K. Usharani, S. M. Jameel Basha, and B. Durga Prasad 32 Voltage Output Characteristics of PZT-5H Disk Under Temperature, Humidity and Traffic Load . . . . . . . . . . . . . . . . . . . . . . . . 329 Dubey Ankush Kumar, Das Biplab, and Gupta Prateek 33 Estimation of SoC Testability at Early RTL Stage . . . . . . . . . . . . . . . . 339 A. Swetha Priya, S. Kamatchi, and E. Lakshmi Prasad 34 Low Power LDO of CMOS Voltage Regulator . . . . . . . . . . . . . . . . . . . 369 B. Naga Lakshmi, B. Suma, and Dayadi Lakshmaiah 35 Fire and Gas Detection System for Home and Industry Safety Using ARM7 Through ESP8266 Wi-Fi Module . . . . . . . . . . . . . . . . . . . 381 M. Sambasiva Reddy, M. Gopi, M. Syam Kumar, and M. Nikita 36 A Relevant Image Recommendation Based on Social Images . . . . . . 391 A. Rajalakshmi, K. S. Riya, and C. Thirumalai Selvan 37 Partial CMMI V2.0 Assessment Using Multi-grade Fuzzy for Healthcare and Insurance Segment in Software Services . . . . . . . 401 Saravana Kumari Sundaram and M. Suresh 38 Automatic Welding on Horizontal Cylindrical Workpiece Using Programmable Logic Controller (PLC) . . . . . . . . . . . . . . . . . . . . 413 Gautam Gupta and Geetika Dua 39 Triple Bottom Line Sustainability and Industry 4.0 Implementation in Indian MSMEs: A Conceptual Model . . . . . . . . . 425 Pavan Rayar and K. N. Vijaya Kumar 40 Bat Algorithm-Based Adaptive Route Change Technique for Pesticide Spraying Using IoT in Dynamic Environments . . . . . . . 433 A. Sherly Alphonse, M. Suguna, S. Abinaya, and D. Jeyabharathi 41 Technical Feasibility of EV Infrastructure with Renewable Power Integration: A Case Study at NIT Srinagar . . . . . . . . . . . . . . . . 441 Zeeshan Hayaat Rather, Sheikh Safiullah, Asadur Rahman, and Shameem Ahmad Lone

Contents

xv

42 An Efficient Topology for Simultaneous Feeding of AC/DC Loads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451 Krishna Kant Dixit and Arti Badhoutiya 43 Utilization of Supplementary Cementitious Material and Waste Marble Powder in Cement and Concrete for Sustainable Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 461 Sharma Neha, Singh Abhishek Kumar, Sharma Prashant, and Parashar Arun Kumar 44 CFD Simulation on Optimum Material to Design and Fabricate the Counter Flow Vortex Tube . . . . . . . . . . . . . . . . . . . . 471 K. Kiran Kumar Rao, A. Ramesh, and Chennabasappa Hampali 45 A Compact Wideband Rectangular Patch Antenna for Wireless Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499 Prathipati Rakesh Kumar, Bondili Siva Hari Prasad, Kudumula Srilatha, and Chopparapu Yogendra 46 Development of a Preliminary Automated CAD Modeling System for Different Types of Rivets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517 Sarikonda Tarun Kumar, Nikhila Pulloju, and Jayakiran Reddy Esanakula 47 The Mechanical Behavior and Tribological Characteristics of Nylon-CaSO4 Polymer Composites . . . . . . . . . . . . . . . . . . . . . . . . . . . 529 S. Sreenivasulu and A. Chennakesava Reddy 48 CFD Analysis of Heat Transfer Characteristics of Automobile Radiator Using Various Nanofluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 541 A. Manojkumar and D. Senthilkumar 49 Stress Analysis of Two High Rolling Mill Chock for Improving Service Life Using Finite Element Analysis . . . . . . . . . . . . . . . . . . . . . . 553 Laxmikant Patil and M. Dubey 50 Modeling of Grid Integrated Solar PV System Using Shunt Active Power Filter (SAPF) with a PI Controller . . . . . . . . . . . . . . . . . 561 M. Kumudwathi, G. Sreenivasan, and R. Kiranmayi 51 Smart Watch for Panic Alert . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571 P. Venkateswara Rao, V. Hymavathi, L. Pavan Kumar, and Ch. Ugandhar Venkata Sai Krishna 52 Quality Assessment of Ground Water at Andaman and Nicobar Islands District, India—A Review . . . . . . . . . . . . . . . . . . . 581 Shraddha Sharma, Geeta Singh, and Gauhar Mehmood

xvi

Contents

53 Heart Disease Prediction Using Machine Learning . . . . . . . . . . . . . . . 589 Ch. Siva Rama Krishna, M. Vasanthi, K. Hemanth Reddy, and G. Jaswanth 54 Performance Analysis of GGBS and Fly Ash-Based Geopolymer Concrete . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 597 Satish Kr. Jangid, Rohit Choudhary, and Manoj Balotiya 55 Development of an Automated CAD Modeling System for a Four-Stroke Four Cylinder IC Engine Crankshaft . . . . . . . . . . . 605 Pranay Pulipaka, Rajan Bhargav Souda, Navanth Pandrala, and Jayakiran Reddy Esanakula 56 Workability and Compressive Strength of Cement Concrete Prepared Using Al2 O3 Nanoparticles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 617 Kamal Kishore 57 Assessment on Multiple Aspects of Online Book Recommendation Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625 Anusha Penugonda and Muralidhar Pantula 58 LED Driver with an Efficient Interleaved Buck Converter . . . . . . . . 635 Srinivas Punna, Sujatha Banka, Guru Sai Charani Ankisetty, Sravani Alluri, Manogna Boora, and Raghamala Vallabhadasu 59 A Comparative Analysis of Customer Behavior Using Sophisticated Machine Learning Techniques . . . . . . . . . . . . . . . . . . . . . 649 K. Nikitha and K. M. Rayudu 60 Object Detection with Voice Feedback for Blind People Using Computer Vision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 661 M. Suguna, Sherly Alphonse, A. Sheik Abdullah, C. U. Omkumar, and D. Prakash Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 673

About the Editors

Dr. Amaranadha Reddy Manchuri is a Ph.D. in Mechanical and Manufacturing Engineering, currently working as a professor and HOD of Mechanical Engineering at MRCET Campus. He is a life member of professional associations ORSI, ISTAM, IndACM, ISTE, EWB, SAE India, and ISSMO. He has received many academic excellence and research innovation awards like the Malaysian Government International Student Award “Malaysian International Scholarship”. His research interests include Mathematical Modelling, DOE and TRIZ, Bioenergy, Pyrolysis of Biomass, Synthesis of Nanomaterials, Engine Performance Analytics, Applied and Fluid mechanics, Spectrophotometry. Dr. Reddy, as a principle investigator, has successfully completed AICTE, Government of India-sponsored research project “Multiobjective Optimization of Production Process Parameters Using Evolutionary Algorithms” and actively guiding Ph.D., PG, and UG research projects. Dr. Amaranadha Reddy Manchuri is the organizing chair and the editor for Springer book series SIST “Proceedings of Intelligent Manufacturing and Energy Sustainability”. Dr. Deepak Marla is currently working as an assistant professor in the Department of Mechanical Engineering at the Indian Institute of Technology Bombay (IIT Bombay). He has obtained Ph.D. from IIT Bombay and had done his postdoctoral work from the Technical University of Denmark and University of Illinois at Urbana-Champaign. His work is in the domain of micro/nano-manufacturing using advanced techniques that involve lasers, electric discharges, electrochemical reactions, plasmas, and micro-tools. His research focuses on gaining a fundamental insight into these processes through a synergetic use of multi-physics modeling and simulation and experiments with an eye on addressing critical challenges at the process level. Dr. Deepak has also bagged couple of prestigious R&D reputable honors and works such as Early-Career Research Award from DST—SERB for laserinduced forward transfer (LIFT) for microscale printing of metals and also a research project on Laser-Assisted Machining of Titanium Alloys from ISRO. Dr. V. Vasudeva Rao received his doctorate from the prestigious IISC, Bangalore. He is presently holding the position of the professor and the chairperson of the School xvii

xviii

About the Editors

Research Committee in the Department of Mechanical Engineering at University of South Africa. He has expertise in Contact Heat Transfer, Nano-Thermofluids, Heat Exchangers, Jet Impingement Cooling, Thermal Management, Energy Systems, and Thermo-Physical properties. He has the prestigious rating with NRF: C2 for six years (2017–2022). Dr. V. Vasudeva Rao has also been an active life member for many technical societies for knowledge empowerment. He has also guided many students for their doctoral and master thesis. He actively publishes technical work on a regular basis which has made him reach with Google Scholar citations of 358 with h-index 10.

Chapter 1

A Hybrid Optimal Technique for Road Extraction Using Entropy Rate Super-Pixel Segmentation and Probabilistic Neural Networks D. Subhashini and V. B. S. Srilatha Indira Dutt Abstract Road extraction is the leading type of application in the area of remote sensing image systems. The major challenges in road network extraction are classified under two categories: intensity and width. The majority of relevant research has focused on machine learning-based methods; however, they have not been able to achieve the highest level of extraction accuracy. Thus, research work is implemented with the advanced deep learning-based probabilistic neural networks classification mechanism to overcome this. Initially, the entropy rate super-pixel segmentation approach is used to efficiently detect the lanes of road, respectively. Finally, to achieve this, probabilistic neural networks were developed for the classification of the road and non-road classes. Texture features, discrete wavelet transform-based low-level features, and statistical color features each have their own co-occurrence matrix at the gray level. Finally, road object analysis operation was performed on the classified outputs; simulation analysis shows that the proposed method shows better qualitative and quantitative analysis.

1.1 Introduction In the study of natural resources and the environment, remotely sensed data [1] is becoming an increasingly major element. Forestry management, crop monitoring, land use monitoring and management, water resources management, urban planning and development, traffic management, natural disaster recovery, military observation, soils mapping, archaeological investigations, mineral prospecting, ocean resources, climate change analysis, deforestation, and so on are just a couple of small applications [2] of remotely sensed data. Road extraction [3] could be used in transportation D. Subhashini GITAM (Deemed to be University), Visakapatnam 530045, India e-mail: [email protected]; [email protected] V. B. S. Srilatha Indira Dutt (B) Department of ECE, GITAM (Deemed to be University), Visakapatnam 530045, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_1

1

2

D. Subhashini and V. B. S. Srilatha Indira Dutt

to conduct traffic studies, assist in the planning of new highways, and discover alternate routes for emergency vehicles [4]. There is an immediate need to find short and safe highways in the event of disasters [5] such as earthquakes, floods, and fires. All of the above-mentioned applications relied on manual based on satellite images to extract road networks. The latter systems for automatic and semi-automatic road extraction have their own set of difficulties [6], including the inability to distinguish roads of various shapes, the inability to map roads with vegetation cover, and partial mapping due to occlusions. The fundamental issue [7] with road extraction is that the difference between road intensity (spectral or color) values varies substantially. Other surrounding things such as vehicles, buildings, trees, and shadows can completely or partially obscure roads, or they can overlap themselves. In comparison with rural areas, the challenge in urban and sub-urban areas is the complex scene content and diverse architecture of the road network. Houses, trees, buildings, autos, parks, and other urban photos are included. The roads do not have a definite line shape, and every road in the city is connected to at least one major route. The next issue with road extraction [5] is that some roads only have one lane in the middle, while others have lanes on both sides. Multiple lanes and zebra crossings may be present on 200-foot highways. Lanes are usually depicted as straight lines, dashed lines, or a combination of the two. This study addresses these concerns and proposes a novel semi-automated approach for extracting road network. The majority of automatic road detections employ an ad-hoc multistage technique. Many scholars have worked on automatic and semi-automatic ways over the last two decades. Automatic road network extraction is difficult to achieve due to the enormous complexity of metropolitan road structures. Many writers have attempted to extract roads from remote sensing photos using multi-resolution-based techniques. The three types of road extraction, detection, and recognition approaches are manual, semi-automatic, and automatic. Human experts identify roads, and with the help of tools, such as mapping, they may readily update transportation databases. The strength, texture, color, and intensity values of the reflecting signals are used to categorize objects into different groups. Combining segmentation and classification [8] to increase road extraction performance is known as combined classification and segmentation. The input image’s morphological operation is sensitive to specific forms. The morphological operation looks for a road with a predetermined size and shape, as well as the surrounding area. Prior information is essential for doing road extraction using the road’s most important property, its intensity [9]. In [10], authors presented a framework to remove disturbances in the image such as objects which has similar spectral characteristics as road surfaces; the morphological trivial opening is utilized here. But, usually it cannot handle occluded road areas by vehicles, shadows of trees and buildings, and other objects. In [11], authors developed a 2 × 2 mask (four-pixel neighborhood with the median value) to create a pyramid of images gradually, from low-resolution to high-resolution imagery. The morphological detector extracts lines from low-resolution images, and

1 A Hybrid Optimal Technique for Road Extraction Using Entropy Rate …

3

each line is considered to correspond to segmented lanes in the original image, allowing the position and extent of these lanes to be determined. In [12], authors presented a motivating and extensive approach to extract the road with a model pattern through multi-scale analysis. They combined coarse scale with finer scale for detailed information. They started with identifying adaptive global thresholding-based morphological edges (as a pair) in the high-resolution imagery and extracted the road network fully from a low-resolution image, by combining edges with lines. In [13], authors created the road’s hierarchical tree structure by merging roadsides derived from a high-resolution picture (less than 0.5 m) with lines recovered from a lower-resolution image. Parallel edges having homogeneous lanes between them are identified from the increased resolution image by using SVM classifier and a corresponding line in the reduced resolution image. A shape feature extractor technique for road extraction using maximum likelihood classification based on multivariate adaptive regression splines (MARS) technique was proposed by authors in [14] on urban area. Three-stage fuzzy inference system was developed, in which initially unsupervised classification (K-medians algorithm) used to segment the image, and then, clustering was done to extract road class based on characteristics of road lanes. A technique, namely iterative self-organizing data analysis technique classification, was proposed by authors in [15], for linear feature extraction based on geometric features, merging and splitting of clusters based on similarity and dissimilarity among neighborhood pixels. A three-stage learning technique to segregate road clusters from others objects was developed by authors in [16] by using the LS-SVM classification. The first one is parameter learning, to choose suitable parameters for clustering algorithm. The second one is used to select a suitable clustering algorithm: either K-means or Knearest-neighbor. Then, last one to separate road from other non-road clusters by using LS-SVM. To develop, the effective road extraction operation must be fulfill the abovementioned challenges and maintain all the considerations mentioned above. To achieve the goals, the research (article) is contributed as follows: (a) entropy rate super-pixel segmentation (ERS) mechanism is used to identify the road lane from the Gaussian-filtered test image. (b) Network is trained and tested with the GLCMbased texture features, DWT-based low-level features and statistical color features by using the PNN deep learning model. (c) The PNN mechanism effectively classifies the road and non-road classes. Then, final road is extracted by using the morphological object analysis.

4

D. Subhashini and V. B. S. Srilatha Indira Dutt

1.2 Proposed Method The proposed research study focuses on lane recognition from roads and lane categorization into road and non-road classes. Figure 1.1 shows the detailed operation of the road extraction, lane detection, and classification approach.

1.2.1 Database Training and Testing The database is designed using photos from the “geographical information systems (GIS) database (GIS)” Archive. GIS is one of the most comprehensive sets of realtime road photos accessible. All the images presented in dataset are trained using the probabilistic neural networks (PNN) network model with gray-level co-occurrence matrix (GLCM) features, statistical and Bayesian classifier. In addition, a random unknown test road picture is sent into the algorithm for detection and categorization.

1.2.2 Preprocessing Background information, haze, and noise are all included in the query image acquired during the image acquisition process. Preprocessing is required to remove the aforementioned undesired portions. Preprocessing is used to remove unwanted elements from the road image, such as unwanted ambient portions, haze (which includes sounds), labels, tape, and glitches. To minimize the hazing effect, a gray-level optimization approach, dark channel prior-based transmission map estimation, and refinement in pixel- and patch-wise manners are utilized in the preprocessing procedure. Thus, the atmospheric road effects are resolved in each pixel-based patch. Salt-and-pepper noise, Gaussian noise, speckle noise, and Poisson noise all appeared in the RSI photographs. When noise occurs in a picture, the pixels reflect varying intensity levels rather than genuine pixel values. This noise reduction procedure

Fig. 1.1 Road extraction and classification

1 A Hybrid Optimal Technique for Road Extraction Using Entropy Rate …

5

will be effective if the Gaussian filter approach is being used in the early phase of preprocessing. Finally, a fast iterative domain guided image filtering (ID-GIF) approach was developed to obtain smoothen output with de-noising properties. The analysis of pixels at various scales, sharpening and smoothing filter de-noising strives to remove the noise exhibited in the pixel, as it conserves the image uniqueness, despite its pixel satisfied, reduces noise to a large extent and minimizes the introduction of visual artifacts. These Gaussian filters successfully identify and remove noise and thin artifacts from the picture; we next use a top-hat transform to remove thick artifacts. The road area is also made subject to contrast-limited adaptive histogram equalization (CLAHE) to get an upgraded picture in the spatial domain. CLAHE operates on the entire image, boosting the contrast; it splits the image and works on the little lanes known as tiles as well. Each tile is generally 8 * 8 pixels, and the histogram inside each tile is equalized, accentuating the area’s edges. Contrast limiting is used to keep the contrast below a certain level in order to reduce noise.

1.2.3 Image Segmentation Following the preprocessing step, lane segmentation was performed to obtain the transparent section of the road area. For segmentation, the entropy rate super-pixel (ERS) segmentation approach was created, which falls within the supervised segmentation technique category. This is a hybrid strategy that uses both regions and thresholding. It is not primarily based on the histogram notion. Super-pixel areas with a high entropy rate are a two-dimensional plane that divides the item from the backdrop. The transition extraction of Eagle picture from its original gray image as an example. The “foreground and background” are depicted in black, while the transition zones are shown in white. The entropy rate probabilistic-pixel zones literally distinguish the foreground from the background. The three characteristics of the initial entropy rate probabilistic-pixel area are: (a) RSI’s regions property: Entropy rate probabilistic-pixel regions contain multiple pixels width at non-step edges, but only one pixel width near step edges. (b) RSI’s boundary property: It is situated “between object and backdrop” and encircles the item. (c) Gray-level variation: Gray levels in transition zones typically shift frequently and intensely, providing a wealth of information for characterizing them. Effective average gradient of entropies is the basis for the approach devised. Traditional gradient-based techniques, on the other hand, are susceptible to noise and are best suited to abrupt sharp level changes rather than periodic gray-level changes. The entropy rate probabilistic-pixel segmentation was used to build several local statistics-based techniques to address this restriction. Transition areas were extracted using local region characteristics including local entropy, modified local entropy, and gray-level difference. Gradient-based techniques to transition area extraction work well for abrupt gray-level shifts.

6

D. Subhashini and V. B. S. Srilatha Indira Dutt

However, it is ineffective in situations when the gray level fluctuates often. Gradient-based techniques are not suitable for pictures with regular gray-level changes rather than rapid gray-level changes. As a result, they presented the content local entropy information measure as a solution to frequent gray-level shifts. The local entropy in a nm local neighborhood for a road picture of size NM is defined as H =−

L−1 ∑

Pi . log Pi

(1.1)

i=0

Pi =

ni n×m

(1.2)

Pi is the probability associated with the RSI ith picture, and H is the local entropy. The number of super pixels with gray levels in HSI is ni , and the highest gray level is L. According to Eq. 1.1, the local variance is higher for heterogeneous lanes and lower for homogeneous lanes. The entropy threshold may be calculated using Eq. 1.3 as follows: E th = α Hmax

(1.3)

where H max is the entropy image’s highest entropy and is a coefficient with a value ranging from 0 to 1. The average values of for sufficient pixel extraction for transition lanes are between 0.8 and 0.9. The following is a summary of the full algorithm: Steps for ERS segmentation (a) For a definite neighborhood, compute the local entropy of input image (b) Extract transition lanes using entropy threshold (c) Find segmentation threshold from the mean of transition lanes histogram (d) Segment the image by comparing it with segmentation threshold (e) Extraction of lanes was identified from the road images

1.2.4 Feature Extraction To categorize the road and non-road classes, several characteristics may be derived from the road picture. Gray-level co-occurrence matrix (GLCM)-based texture features, discrete wavelet transform (DWT)-based low-level features, and statistical color features are some of the main characteristics that assist us separate the road lanes. With the help of the spatial relationship between image pixels, the texture method GLCM analyzes textures. GLCM functions characterize picture texture by computing how many pairs of pixels with specified values and in a certain spatial link are

1 A Hybrid Optimal Technique for Road Extraction Using Entropy Rate …

7

Fig. 1.2 Orientations and distance to compute GLCM

present in the image. After creating a GLCM matrix, statistical texture characteristics may be retrieved from the GLCM matrix. GLCM depicts the presence of various combinations of pixel brightness values, often known as gray levels, in an image. It specifies the likelihood of a given gray level appearing in the vicinity of other gray levels. The GLCM is extracted from the image first in this work for all three color spaces: RGB, CIE L * u * v, and YCbCr. The GLCM matrix is then computed in four directions, as illustrated in Fig. 1.2, 135°, 90°, 45°, and 0° degrees. Let a, b be the number of rows and columns of the matrix, S(a, b) be the probability value recorded for the cell (a, b), and N be the number of gray levels in the picture in the following calculations. Several textural properties can then be derived from these matrices; the extracted textural features are shown in the equations below: GLCM features used are Contrast =

N −1 ∑

Sa,b (a − b)2

(1.4)

a,b=0

Homogeneity =

N −1 ∑

sa,b 1 + (a − b)2 a,b=0

(1.5)

⎤ − μ − μ )(b ) (a a b Correlation = Sa,b ⎣ /( )( ) ⎦ 2 2 σa σb a,b=0 N −1 ∑

⎡

Angular Second Moment (ASM) =

N −1 ∑

2 sa,b

(1.6)

(1.7)

a,b=0

Energy =

√ ASM

(1.8)

The two-level DWT is then used to extract the low-level features. The LL1, LH1, HL1, and HH1 bands are created when the DWT is initially applied to the segmented

8

D. Subhashini and V. B. S. Srilatha Indira Dutt

Fig. 1.3 2-level DWT coefficients

output, in that order. Then, entropy, energy, and correlation properties are calculated using the LL band. Then, DWT is carried out once more on the LL output band, producing the outputs LL2, LH2, HL2, and HH2 accordingly. Again, entropy, energy, and correlation features are calculated on the LL2 band, respectively, as shown in Fig. 1.3. And finally, mean and standard deviation-based statistical color features are extracted from the segmented image. They are Mean(μ) =

N 1 ∑ I (i, j ) N 2 i, j=1

/ Standard deviation (σ ) =

∑N

i, j=1 [I (i, N2

(1.9) j ) − μ]2

(1.10)

In statistical classification, the Bayes classifier is a helpful benchmark. is defined as (perhaps based on training data). As a result, evaluating the success of various classification approaches requires this non-negative variable. Then all these features are combined using array concatenation and results the output as hybrid feature matrix.

1.2.5 Road Region Classification Finance, health, engineering, geology, physics, and biology are just a handful of the fields in which neural networks have proven to be effective. From a statistical approach, neural networks are attractive because of their potential applicability in prediction and classification problems. PNN is a technology that was developed by simulating the brain anatomy of newborns. The neurons are connected in a certain pattern to efficiently perform the categorization task. The hybrid qualities are used to generate the neuron weights. The relationships between weights are then established using its hybrid properties. The number of weights determines the layer levels in the proposed network.

1 A Hybrid Optimal Technique for Road Extraction Using Entropy Rate …

9

Fig. 1.4 Layered architecture of PNN model

Figure 1.4 depicts the architecture of artificial neural networks. PNN divides the categorization process into two stages: training and testing. A layer-based architecture will be used to carry out the training procedure. The input layer is used to map the input dataset, which is separated into weight distributions based on its hybrid features. There are four weighted hidden layers in the PNN design. Following the class node activation layer and the decision normalization layer, the net’s initial convolutional 2D hidden layer accepts 224 * 224 * 3 pixel road segmented photos and applies 961,111 filters at stride 4 pixels. The classification procedure was then implemented in the two tiers of hidden layer class nodes. The two levels of the hidden layer store the normalcy and anomalies of the road extraction characteristic data separately. Based on segmentation criteria, it is regarded as either normal or abnormal categorization stage. These two layers are labeled in the output layer. The non-road extraction kinds are likewise stored individually in the hidden layer, as are the classification weights in the second stage of the hidden layer. These categorization weights are translated as labels onto the output layer in the same way. When the test picture is used, its hybrid features are retrieved and used in the classification step for testing. PNN will begin the classification process based on the maximum feature matching criterion using Euclidean distance. The image is classed as normal road if the feature match happened with hidden layer-C1-labels. It is classed as non-road class if the feature match happened with hidden layer-C2 labels with highest weight distribution.

1.2.6 Road Object Analysis Morphological procedures are performed on the ERS segmented picture using the road classes acquired. If the road is existent, morphological opening is performed using the DWT-GLCM shape characteristics. If the road is not seen, the shape-based statistical color characteristics will be used to delete it. As a consequence, correct road lanes improve the quality of the final photograph.

10

D. Subhashini and V. B. S. Srilatha Indira Dutt

Steps for road object analysis 1. Calculate the upper area limit (AU) and lower area limit (AL) on segmented image 2. Start the connected component labeling procedure for identifying the continuity of the roads by using the PNN output classes 3. If AU > continuity > AL Road presented-retain the object else Remove the object 4. Repeat the procedure for all connected components

1.3 Experimentation Environment 1.3.1 Subjective Analysis The experiments are done using the MATLAB programming language, and road extraction is done using the MATLAb R2018a tool. The PNN architecture was trained with 500 epochs using photos from the GIS dataset for each label, while the remaining 20% was used for testing. The features retrieved by the GLCM and DWT feature networks are utilized to train the PNN classifier, which then classifies the images into their appropriate classes. Various performance measures can be used to determine the model’s efficiency. From Fig. 1.5, it is observed that the proposed method can be effectively detecting the lanes of road RSI images; it indicates the segmentation done very effectively compared to the state of art approaches. Here, the RSI-1 test sample belongs to the semi-urban area, RSI-2 test sample belongs to the sub-urban area, RSI-3 test sample belongs to the developed urban area, and RSI-4 test sample belongs to the Developed urban area with grassland. ERS segmented outcome consisting of road bitmaps with perfect threshold levels. In the PNN classified output, black-color background indicates the non-road class and white color background indicates the road class, respectively. By using these classes, final lane is extracted and road segmented outcome is generated.

1.3.2 Quantitative Analysis For performing the quantitative analysis, mainly accurate classification of road and non-road classes is considered. This classification scenario is calculated by using the True positive (TP), False positive (FP), False Negative (FN), and True Negative (TN), respectively. The GIMP quantitative analysis model is considered as shown in Fig. 1.6 for effective estimation of RSI images. Here, the PNN classified output

1 A Hybrid Optimal Technique for Road Extraction Using Entropy Rate … ERS segmented output

PNN classified output

Final Road extracted output

RSI-4

RSI-3

RSI-2

RSI-1

Test Input image

11

Fig. 1.5 Segmented output images of various methods

image is compared with the reference ground truth image and resulted in estimation of various parameters Quality (E equal ), correctness (E corr ) and completeness (E comp ), respectively. If the completeness is high, it means all the roads are extracted. If the correctness is high, it means roads extracted are original and actual roads, and there is no non-road classes. If the quality is high, it means roads are extracted accurately. E comp =

TP TP + FN

(1.11)

E corr =

TP TP + FP

(1.12)

TP TP + FP + FN

(1.13)

E qual =

12

D. Subhashini and V. B. S. Srilatha Indira Dutt

Fig. 1.6 Evaluation method

Table 1.1 Performance comparison of RSI test cases using the metrics Ecomp, Ecorr and Eequal Metric

RSI-1

RSI-2

RSI-3

RSI-4

E comp

0.9657

0.98099

0.95796

0.97765

E corr

0.97588

0.95024

0.9666

0.98857

E equal

0.98207

0.98494

0.95395

0.97602

Table 1.1 presents the performance comparison of various RSI test cases by using the parameters E comp , E corr , and E equal , respectively. From the table, it is observed for all test images, the PNN classification results in the above 95% accuracy of quality, correctness, and completeness. From Table 1.2, it is observed that the proposed method gives the highest road extraction accuracy for all types of urban areas compared to the conventional approaches respectively. Especially, from Table 1.2a–c, it is observed that the proposed method gives highest accuracy of road extraction in semi-urban, suburban, and developed urban areas compared to the conventional approaches Liu et al. [10], Miao et al. [15], Chaudhuri et al. [11], Singh and Garg [12], and Pramod Kumar et al. [16]. Similarly, from Table 1.2d, it is observed that the proposed method gives the better results in developed urban areas with grass land environment compared to Huang and Zhang [13], Miao et al. [14, 15] and Pramod Kumar et al. [16]; it is achieved because, the proposed approach utilizing the deep learning-based PNN classification, whereas conventional approaches utilizing the simple thresholding methods, adaptive segmentations, shape-based feature extractions and machine learning-based SVM classification. Thus, the PNN has the highest robustness to classify the road and non-road classes, finally resulting in better extraction.

1.4 Conclusion Using a PNN-based deep learning strategy, this paper offered a computational framework for lane recognition and classification of RSI pictures. For preprocessing, Gaussian filters are used, which remove any undesirable noise components or artifacts formed during picture acquisition at the satellites. Then, for ROI-based lane extraction and identification of road area, ERS segmentation is used. Then, for extracting

1 A Hybrid Optimal Technique for Road Extraction Using Entropy Rate …

13

Table 1.2 Performance comparison of various test cases from literature Authors

E comp

E corr

E equal

(a) Semi-urban area with RSI-1 Liu et al. [10]

89.4

73.1

67.3

Miao et al. [15]

90.4

72.8

67.6

Chaudhuri et al. [11]

88.7

70.5

64.6

Singh and Garg [12]

95.32

96.52

92.15

Pramod Kumar et al. [16]

96.4

96.50

92.82

Proposed method

96.57

97.588

98.207

(b) Sub-urban area with RSI-2 Liu et al. [10]

92.1

76.2

71.6

Miao et al. [15]

88.3

79.8

72.2

Chaudhuri et al. [11]

91.6

63.0

66.2

Singh and Garg [12]

–

–

–

Pramod Kumar et al. [16]

76.23

91.99

71.48

Proposed method

98.099

95.024

98.494

64.5

46.4

(c) Developed urban area with RSI-3 Liu et al. [10]

62.3

Miao et al. [15]

55.4

68.3

44.1

Chaudhuri et al. [11]

61.7

62.4

31.0

Singh and Garg [12]

–

–

–

Pramod Kumar et al. [16]

87.39

74.9

67.6

Proposed method

95.796

96.66

95.395

(d) Developed urban grass land area with RSI-4 Huang and Zhang [13]

87.2

78.5

70.4

Miao et al. [14]

68.1

90.2

63.4

Miao et al. [15]

85.7

97.2

83.6

Pramod Kumar et al. [16]

80.15

85.15

70.59

Proposed method

97.765

98.857

97.602

statistical, color, and texture information from segmented images, a GLCM, DWTbased technique was created. Finally, using the pre-trained network model, PNN was used to categorize the road and non-road classes. To construct the final extracted road, a linked component labeling-based objective analysis was used.

References 1. Lian, R., et al.: Road extraction methods in high-resolution remote sensing images: A comprehensive review. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 13, 5489–5507 (2020)

14

D. Subhashini and V. B. S. Srilatha Indira Dutt

2. Abdollahi, A., et al.: Deep learning approaches applied to remote sensing datasets for road extraction: a state-of-the-art review. Remote Sens. 12(9), 1444 (2020) 3. Qi, X., et al.: Deep attention and multi-scale networks for accurate remote sensing image segmentation. IEEE Access 8, 146627–146639 (2020) 4. Spolti, A., et al.: Application of U-net and auto-encoder to the road/non-road classification of aerial imagery in urban environments. In: VISIGRAPP (4: VISAPP) (2020) 5. Zhou, M., et al.: BT-RoadNet: a boundary and topologically-aware neural network for road extraction from high-resolution remote sensing imagery. ISPRS J. Photogrammetry Remote Sens 168, 288–306 (2020) 6. Courtial, A., et al.: Exploring the potential of deep learning segmentation for mountain roads generalisation. ISPRS Int. J. Geo-Inf. 9(5), 338 (2020) 7. Hong, M., et al.: A novel FMH model for road extraction from high-resolution remote sensing images in urban areas. Procedia Comput. Sci. 147, 49–55 (2019) 8. Dai, J., et al.: Road extraction from high-resolution satellite images based on multiple descriptors. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 13, 227–240 (2020) 9. Abdollahi, A., Pradhan, B., Shukla, N.: Road extraction from high-resolution orthophoto images using convolutional neural network. J. Indian Soc. Remote Sens. 1–15 (2020) 10. Liu, R., Song, J., et al.: Road centerlines extraction from high resolution images based on an improved directional segmentation and road probability. Neurocomputing 212, 88–95 (2016) 11. Chaudhuri, D., Kushwaha, N.K., Samal, A.: Semiautomated road detection from high resolution satellite images by directional morphological enhancement and segmentation techniques. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 5(5), 1538–1544 (2012) 12. Singh, P.P., Garg, R.D.: Automatic road extraction from high resolution satellite image using adaptive global thresholding and morphological operations. J. Indian Soc. Remote Sens. 41(3), 631–640 (2013) 13. Huang, X., Zhang, L.: Road centreline extraction from highresolution imagery based on multiscale structural features and support vector machines. Int. J. Remote Sens. 30(8), 1977–1987 (2009) 14. Miao, Z., et al.: Road centerline extraction from highresolution imagery based on shape features and multivariate adaptive regression splines. IEEE Geosci. Remote Sens. Lett. 10(3), 583–587 (2013). https://doi.org/10.1109/LGRS.2012.2214761 15. Miao, Z., et al.: An integrated method for urban main-road centerline extraction from optical remotely sensed imagery. IEEE Trans. Geosci. Remote Sens. 52(6), 1–14 (2014) 16. Soni, P.K., Rajpal, N., Mehta, R.: Semiautomatic road extraction framework based on shape features and LS-SVM from high-resolution images. J. Indian Soc. Remote Sens. 1–12 (2020)

Chapter 2

Meta Heuristic-Based Community Detection of Social Network Using Cuckoo with InfoMap Algorithm S. Devi, M. Rajalakshmi, S. Saranya, and J. Shana

Abstract Facebook and Twitter are one of the social networks for communicating our thoughts, feelings, opinions of current scenarios. By using this communication platform rapidly, people can share information. In the analysis of the social network, identifying such communities and sharing information are a challenging task. These fast shares of information create scalability, inaccurate, genuine issues. Many existing techniques can be used for the community detection of the social network. However, these techniques are ineffective and high time-consuming for quick detection. This paper proposed a novel metaheuristic technique-based community detection of the social network by using the hybrid of cuckoo search with InfoMap algorithm (CIMA). In this proposed work, CIMA is used for community detection. This proposed work increases the intracommunity of the social network and improves its performance in an accurate way by detecting the community of the social network. The accuracy rate for AKSHO got 48%, ant lion got 52%, lion optimization got 66%, fish swarm optimization algorithm got 78%, and our proposed work CIMA got 94%.

S. Devi (B) · M. Rajalakshmi Department of Information Technology, Coimbatore Institute of Technology, Coimbatore, India e-mail: [email protected] M. Rajalakshmi e-mail: [email protected] S. Saranya Department of Computer Science and Engineering, Coimbatore Institute of Technology, Coimbatore, India e-mail: [email protected] J. Shana Department of Artificial Intelligence and Machine Learning, Coimbatore Institute of Technology, Coimbatore, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_2

15

16

S. Devi et al.

2.1 Introduction Analysis the social network expansion with followers and communications between the followers which indicates the complex structure of tweet data [1, 2]. The social network is a combination of various communities, and it is based on various followers. This community is detected by using the hybrid intelligence of the algorithm [3]. In general, a network is a combination of vertices and edges, where edges represent linking between the vertices. The structure of the network can be represented as a graph in which the vertex associates with neighbor vertex and shares the information with it [4]. The social network can be classified into social communities based on individual followers being socially connected with other followers in the concern community [5, 6]. In Twitter, dataset hashtags are an important part of the tweet message through which can group the message of the same topic in a single community; this hashtag usage of tweet dataset is called as hashtag recommendation [7–12]. Community detection is one of the main issues in recent research works. Community detection is based on the NP-hard problem [13]. Traditional methods used for community detection are graph partitioning and hierarchical clustering. Before applying the graph partitioning method first, identify the community number and community size and then partition the network structure. The hierarchical clustering of community detection includes divisive clustering and agglomerative clustering algorithm [14, 15]. The main issues in existing algorithms are detecting community complex structure which is not an efficient one. The detection rate is also high. To overcome this issue, this paper proposes hybrid technique of cuckoo search with InfoMap algorithm in the detection of the social network. By using this method, it can detect complex network structure of community detection in an effective way. In recent years, a lot of research work has been implemented to develop new techniques and algorithms in the detection of community in the social network. One of the most common quality functions is based on modularity as stopping criteria is introduced by Newman. This modularity-based techniques include simulated annealing [16], modularity optimization external optimization [17] and spectral optimization [18]. These modularity-based techniques provide outstanding performance in the detection of community in the social network.

2.2 Related Work The structure of community indicates that group of nodes such that nodes within a community are interconnected with each other and sharing the information through social network [19]. In the analysis of network structure, the detection of community plays a vital role in solving complex network structure. In general, social network communities are linked with each other; this will lead to overlapping structure format.

2 Meta Heuristic-Based Community Detection of Social Network Using …

17

A follower or user may be in more than one community for promoting their business, opinions, sharing information. In the social network, users from various fields try to share their valuable information coming from different fields such as education, politics, marketing, statistics and economy. This paper presented a community detection using deep learning algorithm which includes the concept of deep neural network, graph embedding and graph neural networks. The probabilistic model based on graphical structure is used in the detection of community.

2.3 Proposed Cuckoo Search with InfoMap Algorithm Sentiment community detection of social network using meta heuristic optimization algorithm has been developed with novel metaheuristic technique based on the hybrid of cuckoo search with InfoMap algorithm (CIMA). It includes three stages, namely pre-processing, construct a network community based on influencer, and detecting the community based on hashtag with metaheuristic optimization algorithm of the cuckoo—InfoMap algorithm. This is shown in Fig. 2.1. Figure 2.1 describes that pre-processing stage includes removal of stop words removal, garbage removal, tokenization, etc.

2.3.1 Pre-processing In this stage, it removes unwanted characters like URL, links, web addresses, emoji from the tweet data set. Breaking of tweet streaming data into text, words or meaningful single elements is called tokenization. The process of reducing word from the root into single entity word is called stemming.

Fig. 2.1 Architecture of CIMA

18 Table 2.1 Algorithm 1: construction of network community

S. Devi et al. Construction of network community Input: Number of nodes or vertices n 1 , n 2 , . . . , n m Output: G = (V, E) Step 1: Initialize edges E ← ∅ Step 2: Read the tweet dataset td1 , td2 , . . . , tdm Step 3: Pre-processing using 3.1 Step 4: For each influencer infi ∈ INFL Step 5: Transformation of pre-processed tweet dataset using TF-IDF vectorizer Step 6: tweet_trans ← TF − IDF(tdm ) Step 7: End For Step 8: FOR infi /= inf j ∈ V do

)) ( ( Step 9: IF sim(tweet_trans(infi ), sim tweet_trans inf j > thu then {( )} Step 10: E ← E ∪ infi , inf j Step 11: End IF Step 12: End For Step 13: Return G = (V , E)

2.3.2 Construction of Network Community To construct the network is denoted by G = (V , E) where V is the vertex and E is the edges of the graph. In the community network vertex in graphic G denotes a user and edge denotes communication between two users. The network size is calculated by number of users in the network p = |V | and number ( ) of links in the graph q = |E|. The structure of network is square matrix B = bi j m×m . This square matrix contains values of {0, 1}. If user is connected with another user, then bi j = 1; otherwise; bi j = 0. The main concepts in this network structure are identifying the influencer in the community and detecting the group of influencer using metaheuristic cuckoo search algorithm with InfoMap algorithm. Table 2.1 describes applying the pre-processing on Twitter streaming dataset. For constructing the network of community transforming pre-processed tweet data using TF-IDF (term frequency–inverse document frequency).

2.3.3 Detection of Community Based on Hash Tag Using CMIA (Proposed) In the proposed work is hybrid of cuckoo search with Infomap algorithm (CIMA) is used and it influence the community with hash tag. This InfoMap algorithm is faster

2 Meta Heuristic-Based Community Detection of Social Network Using …

19

in constructing and detecting communities. In the network G = (V , E), applying the community detection algorithm and detected communities are embedding into the n n in the network G. Here, f i=1 comi ⊆ V . In the community comi , network {comi }i=1 the follower f will be included and f must post at least one tweet with hashtag. If follower belongs to more than one community, it means intersection of communities is non-empty.

2.3.3.1

Cuckoo Search Algorithm

Table 2.2 describes that meta heuristic cuckoo search algorithm. By nature, it randomly chooses the nest for laying eggs. In order to implement it, parasitic habit follows three rules. Rule 1: In the randomly chosen nest cuckoo lay one egg at a time. Rule 2: In order to move over next generation, nest with high quality of eggs is selected. Rule 3: Fix the number of available host nest and egg laid by a cuckoo identified by the host bird and the probability pbc ∈ [0, 1]. If alien egg is discovered by the host bird, either it can throw away the egg or discard the nest and built a new nest. That is worst eggs are replaced by the new egg. Table 2.2 describes the solution is based on egg in the nest and replacing the new solution is defined by cuckoo egg. The main aim of this algorithm is replacing the worst solution by new one [20, 21]. Table 2.2 Algorithm 2: cuckoo search algorithm

Cuckoo search algorithm Begin Objective function f n (x), x = (x1 , x2 , . . . , xn )T Create initial population of n host nests, xi , i = 1, 2, 3, . . . , n count = 1 While count ≤ stopcriteria or Max_Iteration do Randomly choose cuckoo xi using levy flights and it is considered as new solution xnew Randomly select the nest x j and calculate the fitness for fit(xnew ) ( ) IF fit(xnew ) > fit x j then Replace x j by new solution xnew End IF ( ) Probability proba of worst nest is abandoned and built a new one compare the worst nest with new one and select the better one Ranking solution and choose the current best end while End

20

S. Devi et al.

Fig. 2.2 Architecture of CIMA

2.3.3.2

InfoMap

In the network of community, InfoMap algorithm starts with each node, and its community then repeats it and merges two communities into one which reduce the code length.

2.3.3.3

Hybrid of Cuckoo Search with InfoMap

In order to get fined detection of community network, this hybrid cuckoo search with InfoMap algorithm is implemented (CIMA). The architecture of CIMA is given in Fig. 2.2. After pre-processing, the tweet data constructs the network community using Algorithm 1. For detecting the communities of network using CIMA based on hash tag, ranking the hash tag is based on the tweet’s similarity score and followers in the community.

2.4 Results and Discussion 2.4.1 Evaluation of Performance The detection of community in the social network is analyzed by two criteria like accuracy and normalized mutual information (NMI) of community detection. Clustering of users in communities based on social network in the tweet data set. The performance metric measures are True Positive (TP), False Positive (FP), True Negative (TN), and False Negative (FN) True Positive (TP): Users or followers have been selected as community members, and their features in the tweet dataset are same to each other. Figure 2.3 shows that NMI criteria of detection of community in the social network of proposed method (CIMA) are high for the detection of communities. The majority of followers from each community are having same followers in the social network. The proposed work CMIA is compared with existing algorithms like

2 Meta Heuristic-Based Community Detection of Social Network Using … Fig. 2.3 NMI criteria for proposed work

21

Accuracy Rate

NMI of Community Detection 1.5 1 0.5 0 1

2

3

4

5

6

Communities

Table 2.3 Performance of metric measures in the aspects of precision, recall and specificity Method name

Precision

Recall

Specificity

AKHSO

93.6 ± 0.6

96.3 ± 0.7

97.3 ± 0.4

Ant lion algorithm

97.2 ± 0.7

95.9 ± 0.3

98.6 ± 0.2

Lion optimization algorithm

94.3 ± 0.5

94.9 ± 0.4

96.6 ± 0.3

Fish swarm optimization algorithm

91.3 ± 0.8

93.9 ± 0.3

95.6 ± 0.4

CIMA (proposed)

98.2 ± 0.8

97.5 ± 0.2

99.4 ± 0.3

discrete Krill Herd swarm optimization (AKHSO) [22], ant lion algorithm [23], lion optimization algorithm [24], fish swarm optimization algorithm [25]. Table 2.3 shows performance of metric measures in the aspects of precision, recall and specificity. In a comparative analysis of cuckoo search algorithm values for existing community search algorithms like AKHSO, ant lion algorithm, lion OA, fish swarm optimization algorithm for same data sets revealed that the proposed algorithm fives better output as refer Table 2.3. The proposed work (CIMA) is focused on percentage of community detection with high average hit rate performance, which is better than other existing algorithms. It is 6.70% of communities achieving higher than 0.9 average hit rate.

2.5 Conclusion Heuristic approach based on community detection of the social network using cuckoo with InfoMap algorithm is implemented. This work collected the tweet data from the streaming data, and it underwent pre-processing. Based on the pre-processed tweet data, we constructed the network, and detection of community is carried out by hybrid algorithm of cuckoo search with InfoMap algorithm. Community-based hashtag framework detects the hashtag and shared with other community followers. In the performance analysis, our proposed work shows that detection of community based on hash tag is better with TF-IDF features, the accuracy rate for AKSHO got 48%, ant lion got 52%, lion optimization got 66%, fish swarm optimization algorithm

22

S. Devi et al.

got 78%, and our proposed work CIMA got 94%. In future, this work is extended up to detection of community which is based on trust based, topic n = based with various similarity measures.

References 1. Messaoudi, I., Kamel, N.: A multi-objective bat algorithm for community detection on dynamic social networks. Appl. Intell. 49(6), 2119–2136 (2019) 2. Tavakoli, S., Hajibagheri, A., Sukthankar, G.: Learning social graph topologies using generative adversarial neural networks. In: Proceedings of the International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction, São Paulo, Brazil (2017) 3. Chand, S., Mehta, S.: Community detection using nature inspired algorithm. In: Hybrid Intelligence for Social Networks, pp. 47–76. Springer, Germany (2017) 4. Said, A., Abbasi, R.A., Maqbool, O., Daud, A., Aljohani, N.R.: CC-GA: a clustering coefficient based genetic algorithm for detecting communities in social networks. Appl. Soft Comput. 63, 59–70 (2018) 5. Zou, F., Chen, D., Huang, D.-S., Lu, R., Wang, X.: Inverse modelling-based multi-objective evolutionary algorithm with decomposition for community detection in complex networks. Physica A: Stat. Mech. Appl. 513, 662–74 (2019) 6. Pournajaf, L., Tahmasebian, F., Xiong, L., Sunderam, V., Shahabi, C.: Privacy preserving reverse k-nearest neighbor queries. In: Proceedings of the 2018 19th IEEE International Conference on Mobile Data Management (MDM), pp. 177–186. IEEE Computer Society, Aalborg, Denmark (2018) 7. Kowald, D., Pujari, S.C., Lex, E.: Temporal effects on hashtag reuse in Twitter: a cognitiveinspired hashtag recommendation approach. In: Proceedings of the 26th International Conference World Wide Web Geneva, Switzerland, pp. 1401–1410 (2017) 8. Li, J., Xu, H., He, X., Deng, J., Sun, X.: Tweet modeling with LSTM recurrent neural networks for hashtag recommendation. In: Proceedings of the International Joint Conference on Neural Network (IJCNN), pp. 1570–1577 (2016). http://dblp.uni-trier.de/db/conf/ijcnn/ijc nn2016.html 9. Li, T., Wu, Y., Zhang, Y.: Twitter hash tag prediction algorithm. In: Proceedings on the International Conference on Internet Computing, pp. 1–5 (2012) 10. Wang, Y., Qu, J., Liu, J., Chen, J., Huang, Y.: What to tag your microblog: hashtag recommendation based on topic analysis and collaborative filtering. In: Chen, L., Jia, Y., Sellis, T., Liu, G. (eds.) Web Technologies and Applications, pp. 610–618. Springer, Cham (2014) 11. Godin, F., Slavkovikj, V., De Neve, W., Schrauwen, B., Van deWalle, R.: Using topic models for Twitter hashtag recommendation. In: Proceedings of the 22nd International Conference on World Wide Web (WWW Companion), pp. 59–596 (2013) 12. Kou, F.-F., et al.: Hashtag recommendation based on multi-features of microblogs. J. Comput. Sci. Technol. 33(4), 71–726 (2018) 13. Fortunato, S.: Community detection in graphs. Phys. Rep. 486(3), 75–174 (2010) 14. Newman, M.E.J.: Fast algorithm for detecting community structure in networks. Phys. Rev. E—Stat. Nonlin. Soft Matter Phys. 69(6), 5p, Article ID 066133 (2004) 15. Clauset, A., Newman, M.E.J., Moore, C.: Finding community structure in very large networks. Phys. Rev. E 70(6), Article ID 066111 (2004) 16. Guimer‘a, R., Amaral, L.A.N.: Functional cartography of complex metabolic networks. Nature 433(7028), 895–900 (2005) 17. Duch, J., Arenas, A.: Community detection in complex networks using extremal optimization. Phys. Rev. Stat. Nonlin. Soft Matter Phys. 72(2), Article ID 027104 (2005) 18. Newman, M.E.J.: Modularity and community structure in networks. Proc. Natl. Acad. Sci. U.S.A. 103(23), 8577–8582 (2006)

2 Meta Heuristic-Based Community Detection of Social Network Using …

23

19. Fortunato, S.: Community detection in graphs. Phys. Rep. 486, 75–174 (2010) 20. Yang, X.-S., Deb, S.: Cuckoo search via L´evy flights. In: Proceedings of the World Congress on Nature and Biologically Inspired Computing (NABIC ’09), pp. 210–214, Coimbatore, India (2009) 21. Rosvall, M., Bergstrom, C.T.: Maps of random walks on complex networks reveal community structure. Proc. Natl. Acad. Sci. USA 105, 1118–1123 (2008) 22. Ahmed, K., Hafez, A.I., Hassanien, A.E.: A discrete Krill herd optimization algorithm for community detection. Paper presented at the 2015 11th international computer engineering conference (ICENCO) (2015) 23. Babers, R., Ghali, N.I., Hassanien, A.E., Madbouly, N.M.: Optimal community detection approach based on ant lion optimization. Paper presented at the 2015, 11th international computer engineering conference (ICENCO) (2015) 24. Babers, R., Hassanien, A.E., Ghali, N.I.: A nature-inspired metaheuristic lion optimization algorithm for community detection. Paper presented at the 2015 11th international computer engineering conference (2015) 25. Hassan, E.A., Hafez, A.I., Hassanien, A.E., Fahmy, A.A.: Community detection algorithm based on artificial fish swarm optimization. In: Intelligent Systems’ 2014, pp. 509–521. Springer (2015)

Chapter 3

Smart Mirror Using Raspberry Pi 4 Desu Sai Pranav Reddy, Y. Sreevatsal Pranav, Padmavathi Kora, and V. Arvind

Abstract The smart mirror’s design, construction, and operation are described and implemented in this study. Every morning, we start our day by looking in the mirror at least once before leaving our homes. We use it mentally to figure out how we look and what we’re wearing, and sometimes just by instinct. One of the Raspberry Pi’s applications is smart mirror, often known as magic mirror. A computer screen is placed in a mirror that has a futuristic appearance. The Raspberry Pi remains in the background scenes and is in charge of the data displayed on the mirror. You may check numerous notifications from social media sites as well as news, weather forecast, and other things while just staring at a mirror. The Raspberry Pi is connected to the monitor through a HDMI cable and also includes a built-in Wi-Fi and bluetooth interfaces, allowing us to stream music and videos on the screen of the mirror.

3.1 Introduction Mirrors are unavoidable and inevitable in our daily lives, and as such, they are indispensable in homes. The current situation is vastly different from that of the past. Every household contains at least one mirror. When individuals began to place a premium on beauty in their lives, mirrors became increasingly important. It assists us in understanding how we show ourselves and keeping our appearances neat. We are concerned about what others think of us and spend countless hours in front of the mirror. For those who consider a mirror as a decorative piece for their home, it can be atomized using an IoT device and turned into a smart mirror. While a generic mirror can be converted into a smart mirror [1] with the help of a Raspberry Pi device. The Raspberry Pi remains in the background and runs all the necessary applications of the smart mirror. The smart mirror’s goal is to make accessing information services such as news feeds, weather, and the clock simple. It also has some rudimentary functions, such as D. S. P. Reddy (B) · Y. S. Pranav · P. Kora · V. Arvind Department of ECE, GRIET, Hyderabad, Telangana, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_3

25

26

D. S. P. Reddy et al.

real-time user interaction. Thus, we intended to build a smart mirror with the above functionalities using the required tools and the necessary equipment.

3.2 Significance and Simulation The smart mirrors were constructed on the basis of knowledge gained through self assessment keeping public perspectives. Though constructing and configuring a smart mirror is a complex technique but, an effective programming approach can resolve it. The smart mirror will combine technology and a mirror to offer information to consumers while they are using it. The smart mirror’s main goal is to improve people’s quality of life [2]. The majority of technology progress for smartphones and tablets is motivated by providing information to users in the most comfortable way feasible. Every day, the smart mirror will present consumers with useful information on their mirror. Allowing users to multitask by consuming media while getting ready for the day will save time for individuals all over the world. The mirror’s purpose is to supply people with information they might need in the morning as they get ready for the day or at night as they prepare for bed [3]. The inspiration for this project comes from a variety of places. The primary character in the Iron Man films [4] uses holographic displays to accomplish a variety of tasks around the house. Corning produced a video about their Glass product a few years ago, which is designed to provide a smart surface everywhere in the home. While these and a slew of additional instances are beyond the scope of this mirror, their realization appears to be a long way off. The smart mirror has the advantage of being simple to implement, despite the fact that it lacks the complex functionality of these examples.

3.3 Framework and Functions of the System According to the practical use, this paper contains a smart mirror design that is ideal for a variety of circumstances. The mirror adapts to the Raspberry Pi 4 as the control processing center for its functioning [5]. The smart mirror has a large number of modules which include the basic modules, data modules, weather modules, informational modules, and control modules [6] (Fig. 3.1).

3 Smart Mirror Using Raspberry Pi 4

27

Fig. 3.1 Smart mirror functional diagram

3.4 Hardware Configuration The system hardware consists of a two-way mirror, frame, display monitor, mobile phone, Raspberry Pi, heat sink, micro SD card, and power supply. The Raspberry Pi was installed, and an LED screen was installed. A back panel was attached to the frame after connecting the HDMI and power connections. A heatsink was installed on the Raspberry Pi 4 to avoid overheating and assembled the hardware accordingly (Fig. 3.2).

28

D. S. P. Reddy et al.

Fig. 3.2 Smart mirror hardware setup

3.5 Design and Implementation The smart mirror layout was designed and implemented using Javascript [7] and configured with multiple modules for the desired functions. A Basic Modules (a) Date and Time Module: The Date and Time module is one of the Smart Mirror’s default modules. The current date and time are displayed in this module. When linked to the internet, the information will be updated in real time.

(b) Flipping Clock Module: This module provides us with a classic flipping clock. It is a module powered by the flip plugin.

3 Smart Mirror Using Raspberry Pi 4

29

(c) Compliments Module: The compliments module is one of the smart mirror’s default modules. This module shows a compliment at random. The compliment that is to be displayed can be edited by typing it out in the code repository.

B Data Modules (a) Calendar Module: The calendar module displays events from a public.ical calendar [8]. It can combine multiple calendars as well. The following image shows display of calendar in Smart Mirror layout on the Top Left position right below the Date and Time.

(b) Notifications Module: This smart mirror module uses the PushBullet API [9] to display phone notifications. This module can be used to show notifications, send a few commands to your magic mirror, or make your magic mirror speak.

30

D. S. P. Reddy et al.

(c) Gmail Feed Module: This smart mirror module builds a table containing the current unread Gmail mails. Instead of IMAP, this module uses the Gmail RSS feed.

C Weather Modules (a) Current Weather Module: This module will be configurable to be used as a current weather view or to show the forecast. Thus, this module can be used twice to satisfy both purposes.

(b) Weather Forecast Module: This module consists of elements that display the weather forecast for the duration of the upcoming week, which also includes an icTon to display the current weather conditions, the minimum temperature, and the maximum temperature of that particular week.

3 Smart Mirror Using Raspberry Pi 4

31

D Control Modules (a) Pages Module: You can have pages in your smart mirror with this module. Every page can be assigned to contain a specific number of other modules and displays it periodically. The total number of pages can be set while configuring the smart mirror modules. (b) Pages Indicator Module: This module indicates the page number visually. The commands are run at the root of the smart mirror installation. (c) Remote Control Module: This module allows us to instantly shut off your mirror using a web browser. Any device should be able to access the website (desktop, smart phone, tablet, …). You may also conceal and show modules on your mirror, as well as perform other amazing things.

E Informational Modules (a) News Feed Module: The news feed module is one of the smart mirror’s default modules. This module uses an RSS feed to display news headlines. Scrolling through news headlines is timed, but it can also be managed by sending the module news feed-specific notifications.

32

D. S. P. Reddy et al.

(b) Stocks Module: This module is an add-on module that has been specially added to the smart mirror by us. The main function of this module is to display the selected stock listings along with its current market prices and price changes. The data is acquired using Yahoo Finance API [10].

(c) Cryptocurrency Module: The following module is mainly used for displaying the selected, current values of multiple cryptocurrencies along with their symbols or logos on the top right corner of the smart mirror display. The data is acquired using coinbase API [11].

(d) NBA Module: This module is a specially integrated module whose main function is to display the live scores of National Basketball Association (NBA). The data is acquired using NBAs API key [12].

3.6 Conclusion Modern products and electronics are increasingly being developed with built-in “smart” features; however, smart mirrors like the magic mirror have yet to gain traction. The absence of software for developing and setting a smart mirror’s layout is one issue that limits a possible smart mirror product. We wanted to see if a designer

3 Smart Mirror Using Raspberry Pi 4

33

interface prototype could be made to assist users in developing and tweaking a mirror layout for our major project. The smart mirror, which also functions as a smart home control platform, is a future system that will provide clients with an easy-to-use mirror interface that will allow them to engage with tailored services while performing other tasks. The main benefits are that it is a unique form of smart device that isn’t seen very often and that it looks great. The mirror can be used as a standard mirror or as a mirror that shows the authorized users daily notifications. On the mirror, the user can access a range of feeds or notifications, such as Facebook, Gmail, and news. Information such as the time, weather, and date can also be displayed on the mirror. The mirror also functions as a personal assistant, displaying important notifications and functioning as a resource center for information. Smart mirror design is perfect for a number of applications such as college, home, and offices due to its small size, simple operation, low cost, high degree of user friendliness, personalized user interface, and many other features. Overall, the smart mirror system proposed combines a variety of characteristics to provide clients with personalized information offerings.

References 1. Two Way Mirrors: Smart Mirror Product Information. www.twowaymirrors.com/smart-mirror/ 2. Handyman: Smart Mirror Work. https://www.familyhandyman.com/article/what-to-knowabout-smart-mirrors/ 3. Princy, A.J.: Technology and Smart Features. https://www.researchdive.com/blog/technologyand-mirror-to-Go-hand-in-Hand-to-offer-smart-features 4. Mashable: This digital butler could make you feel like Iron Man. https://mashable.com/article/ duo-home-mirror-computer 5. Raspberry Pi: Raspberry Pi 4. https://www.electronicwings.com/raspberry-pi/raspberry-pi-int roduction 6. Magic Mirror 2: Module Development. https://docs.magicmirror.builders 7. Javascript: The Modern JavaScript Tutorial. https://javascript.info 8. iCalendar: public.ical. https://icalendar.org 9. PushBullet: PushBullet API. https://docs.pushbullet.com 10. Yahoo Finance: Yahoo Finance API. https://www.yahoofinanceapi.com 11. CoinBase: CoinBase Digital Crypto Currency API. https://developers.coinbase.com/api/v2 12. NBA: API NBA. https://rapidapi.com/api-sports/api/api-nba

Chapter 4

Garbage Bin Alert System Using Arduino UNO for Smart Cities Gangidi Harathi, Gudise Anusha, G. Vinay Raj, and Sowjanya Ramisetty

Abstract The human population is rapidly expanding, posing a significant challenge to waste management systems and, as a result, to keeping the environment clean and green. Many cities around the world are in jeopardy as a result of this issue. We are implementing a smart dustbin employing IoT devices such as Arduino UNO, ultrasonic sensor, gas sensor, humidity sensor, PIR sensor, servo motor, GSM module, and GPS modem to avoid waste overflow from the dustbin, separation of dry and wet rubbish, and also to stop unsanitary conditions. In our paper, we presented an innovative smart dustbin that sends notifications to responsible authorities when the garbage level is reached or if it generates a strong odour, resulting in a more effective and efficient garbage disposal system as well as the ability to solve the problem of dry and wet garbage separation. When motion is detected inside the defined sensor range, the designed system immediately sends an alarm message to the mobile phone.

4.1 Introduction In India, as the population grows, the amount of wastage increasing. In terms of waste collection, transportation, and disposal, India is undergoing significant environmental changes [1]. As a result, rubbish overflow and foul odour can be decreased using IoT devices [2]. The IoT is defined as the non-human linking of surrounding devices across a wired or wireless network. Multiple bins are put throughout the city or on campus, and each bin is connected to an ultrasonic sensor that helps identify the level of waste in the dustbin. So that it is simple to tell which bin is filled [3]. The GSM module sends the notice, which includes the location of the trashcan for the user. The usage of a GSM module and a sensor solution is presented, which sends the dustbin’s notification to the registered phone number. The Internet of things (IoT) idea has been employed for data exchange, processing, storage, and retrieval in order to make the monitoring system more effective. The proposed effort aids in the elimination of the G. Harathi · G. Anusha · G. V. Raj · S. Ramisetty (B) KG Reddy College of Engineering & Technology, Hyderabad, Telangana, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_4

35

36

G. Harathi et al.

daily challenge of managing rubbish in the environment, which is made possible through the Internet of things [4]. An Arduino UNO and a trashcan are connected to ultrasonic sensors, a gas sensor, a IR sensor, and a soil moisture sensor in this setup. GSM and GPS modules are used to keep track the bin. IR sensor is used for object detection, and soil moisture helps us to detect the dry and wet waste. We can either deposit the waste in government-assigned containers in the area/locality or hand it over to door-to-door garbage collectors, but the garbage must eventually reach its final destination, which is critical [5].

4.1.1 Purpose Municipal employees are amongst the first responders in the fight against infectious diseases. Garbage is overflowing in several regions. As a result, the general public and municipal employees may act as disease carriers. This was one of the primary issues that made control difficult. Contagious diseases are becoming more difficult to control as the world’s population grows exponentially and population density rises. So, to help mitigate this problem to some extent, we developed a smart trashcan that detects garbage levels, foul odours, dry, and wet waste separation and sends a notification to the appropriate authorities, which can assist in keeping the environment clean.

4.1.2 Background This paper discusses how to keep contagious diseases from spreading from infected people to healthy people through contact. We are doing this by putting in place a smart waste collection system. When finished, our prototype will help to prevent rubbish overflow and maintain the environment clean. In addition, the sensors detect the amount of waste in the container as well as dry and wet waste separation. During the COVID-19 epidemic, robotics were widely used in quarantine to reduce the danger of infection to front-line workers. Baharin created a robot that was similar but with a different function. The bot could take the patient’s temperature, administer medicine, and monitor their progress using facial recognition. Furthermore, researchers from India’s Lovely Professional University have built a smart trashcan that can work in a close environment, following a predefined path and collecting rubbish without touching it, using a Raspberry Pi and a Mega 2560 processor. To overcome the constraints of this system, several adjustments are made with the help of IOT devices [6]. The paper main purpose is to minimize garbage levels and bad odours. The ultrasonic sensor is linked to the dustbin’s lid, and it detects the waste level in the bin; the gas sensor is connected to the dustbin’s top, and it detects the bad odour from the dustbin; soil moisture sensor detects the dry and wet waste, and the GSM module operates as a messenger, communicating the dustbin’s status to the proper authorities.

4 Garbage Bin Alert System Using Arduino UNO for Smart Cities

37

4.2 Literature Survey Dustbins are equipped with a wireless system that uses a GSM module to display rubbish status. The main goal of the smart dustbin utilizing Arduino UNO is to prevent rubbish overflow and detect foul odours from the dustbin, all whilst improving the cleanliness of the city. This approach is advanced in terms of automated waste monitoring. IoT is a cutting-edge solution that allows us to track the status of the trash can and contributes to a cleaner environment. Sinha et al. [7] suggested a system that uses sensors to monitor garbage levels and provides notifications to the user to keep them informed about the status of the bin. The three-tiered strategy proposed by Singh and Kaur [8] is divided into two parts. Dustbin layer consists of Internet and Wi-Fi-enabled dustbins with a sensor that detects when the bins are full and sends the data to the server. Narayan Sharma et al. [9] created a system that turns a regular trash can into a smart one by employing sensors to monitor garbage levels and sending messages to the user about the bin’s status. Monika et al. [10] suggested a system that makes a regular trash can smart by employing sensors to monitor garbage levels and sends messages to the user to update the bin’s condition. Subho et al. [11] developed a system that analyzed current dustbins and the populations they serve quantitatively. The study looks at the geographical distribution of dust bins in different parts of Dhaka city using average closest neighbour techniques from a geographic. Information System: A approach for decreasing food waste was proposed by Hong et al. Wireless mesh networks enable battery-powered smart garbage bins to interact with one another in a smart garbage bin, and a router and server collect and analyze data for service delivery. Sahu et al. presented a system in which a camera and a load cell sensor are mounted at the base of each waste container at each garbage pickup station. The garbage container will be photographed continuously by the camera. Tripathi et al. suggested an RFID reader model that leverages RFID tags. The system will next utilize the GSM module to validate the user ID from the server database and open the dustbin lid. The ultrasonic sensor will check the rubbish level before the lid opens, and the mechanism will stop working once the dustbin is full. K. Suresh et al.: Arduino UNO Microcontroller-Based Smart Dustbins For Smart Cities-808 sends a text message to the authority informing them that the trashcan is full and then opens the lid.

4.3 Methodology Ultrasonic sensor, gas sensor, humidity sensor, PIR sensor, IR sensor, and soil sensor are the sensors in this proposed model. Sensor for ultrasonic waves. The waste reflects the electromagnetic signals of the ultrasonic sensor, allowing the level of garbage to be determined in relation

38

G. Harathi et al.

to the dustbin’s height. We employ the “HC SR-O4” sensor in our model because the time delay between transmitting and receiving the signal allows us to calculate the exact level of trash in the bin. Gas Detector. A gas sensor, also known as an odour detector, consists of a sensing element that reacts to the volatile nature of the garbage being collected, producing an electrical output that can be used to determine the predominant odour around the trash can. Sensor for Humidity. A humidity sensor is an electronic device that detects humidity in the environment and converts the information into an electrical signal. Infrared Sensor. An infrared sensor is an electronic device that emits rays in order to detect specific aspects of its surroundings. An infrared sensor can both detect motion and measure the temperature of an object. Almost everything emits some form of infrared heat radiation. PIR Sensor. A human moving within 10 m of an infrared sensor is detected. The actual detection range is between 5 and 12 m; thus, this figure is an average. PIR is powered by a pyroelectric sensor that can detect large amounts of infrared radiation. GSM Module. It is a communication standard that connects a terminal to another GSM system. The GSM system handles all signal transmission and reception from the controller to other devices and from other devices to the controller. Module GPS. Our model’s GPS feature aids in the location of various waste collection stations around a neighbourhood. When the waste in a particular neighbourhood’s dustbin reaches an alarming level, the dustbin’s controller uses the GPS system to transmit a command or warning signal to the proper authorities. Soil Sensor. Two probes make up the soil moisture sensor, which measures the volumetric content of water. The two probes enable current to flow through the soil, providing the resistance value needed to calculate the soil’s moisture content. Figure 4.1 specifies the block diagram of the garbage bin alert system. The ultrasonic sensor, gas sensor, humidity sensor, PIR sensor, GSM module, and GPS modem are all linked to the Arduino UNO in this article, which is where the code is dumped. The waste bin’s state will be sensed by sensors, which will transmit a signal to the Arduino UNO. It sends an SMS to the relevant authorities regarding the state of the trashcan using the GSM module. The acronyms GSM and GPS stand for “global system for mobile communication” and “global positioning system,” respectively. The slave unit’s working objects are an Arduino UNO, an ultrasonic sensor, a potentiometer, and a GSM and GPS SIM908 module. The entire unit is placed on top of the dustbin. The ultrasonic sensor sends the trigger pulse into the bin, and the bin receives the echo pulse as a result. It continuously checks the waste level as a result of this. When the waste level reaches a certain threshold, the Arduino UNO receives a signal from the sensor. GSM stands for global system for mobile communication and was designed by ETSI for the second generation to explain mobile phone protocols. Over 80% of market share, GSM became the fact of global standard for mobile communication. ARM microcontroller is interfaced with the GSM modem.

4 Garbage Bin Alert System Using Arduino UNO for Smart Cities

39

Fig. 4.1 Block diagram of garbage bin alert system

When the waste level is reached, it sends a GSM message to the appropriate authorities. The ARM microcontroller is connected to the MAX-232 GSM module. The GSM module features a slot for a SIM card that sends a message to the user when the bin is full or if a foul odour is detected. GSM is used to transmit and receive SMS. The sensor module will receive the input from the waste bins, which are positioned in various regions. The sensor is set at the maximum level of the bin, and if the waste level is exceeded, the sensor detects it and communicates with the GSM module and Arduino UNO, notifying the appropriate authorities that “the dustbin is full, collect it.” By triggering the specific motor of the lid of the wet or dry portion of the dustbin and opening that lid to allow the garbage to slip in, the humidity sensor separates the waste accordingly. The Arduino UNO board will be designed in such a way that if the humidity measured is greater than a threshold value, it will activate the wet part of the dustbin’s motor; otherwise, it will activate the dry section of the dustbin’s motor. The programme to calibrate the PIR sensor was also put into the Arduino UNO, and the PIR sensor was linked to the node MCU module through the Arduino UNO’s I/O ports. If there is any motion inside the defined sensor range, an alarm message is automatically sent to the mobile phone.

40

G. Harathi et al.

4.4 Results The system was checked several times by adjusting the garbage level in the bin. The dustbin is notified to the appropriate authorities at every level. The user then receives the notification and checks the message on the mobile device seen in Fig. 4.2. The system is implemented in the manner in which we envisaged it. Further steps should be taken based on the results. Figure 4.2 specifies the overall about the project explanation, i.e. checking the dustbin status, bad odour, object detection, dry, and wet waste separation with the help of sensors. Figure 4.3 specifies the message regarding the status of the dustbin, i.e. dustbin is full and along with the location. Figure 4.4 shows the message which is sent to concern authority about the person detection.

Fig. 4.2 Garbage bin alert system using Arduino UNO for smart cities

Fig. 4.3 Receiving alert messages about the dustbin status

4 Garbage Bin Alert System Using Arduino UNO for Smart Cities

41

Fig. 4.4 Message regarding person detection

4.5 Conclusion and Future Scope To summarize, adopting this smart dustbin can minimize waste in our country to the greatest extent possible. Because trash management is becoming a serious concern in India, it has a bright future. Our country is producing over 62 million tonnes of garbage. As a result, adequate waste management procedures must be taken in order to achieve a cleaner and healthier nation. If the waste level or any motion in the dustbin reaches a certain level, indicating that the dustbin is full, a notification is sent to the appropriate authorities via the GSM module, including the exact position of the dustbin. With a gas sensor, it prevents bacterial growth. Additionally, it uses soil moisture sensor automatically to separate wet and dry garbage. We can decrease the growth of bacteria and rubbish overflow using this technique, resulting in a cleaner environment. In the future, we can make dustbin which is automatically movable.

References 1. Reddy, P.S.N., Naik, P.N., Kumar, A.A., Kishore, S.N.K.: Wireless dust bin monitoring and alert system using Arduino. In: Second International Conference on Electrical, Computer and Communication Technologies (ICECCT) 2. Bajaj, A., Reddy, S.: Garbage monitoring system using IOT. Int. J. Pure Appl. Math. 114(12), 115–161 (2017) 3. Smart Dustbin. Int. J. Eng. Adv. Technol. 4. Anitha, A.: Garbage monitoring system using IOT. In: IOP Conference Series: Materials Science and Engineering 5. Christopher, M.P.Jr., Salado, M.J.M.L., Sobejana, N.P.: SPAMAST smart garbage bin monitoring system using wireless sensor network. J. Eng. Res. Rep. 6. Lincy, A.F., Sasikala, T.: Smart dustbin management using IOT and Blynk application. In: Fifth International Conference on Trends in Electronics and Informatics (ICOEI) (2017) 7. Kumar, M., Sinha, T., Saisharan, P.: Smart dustbin. Int. J. Ind. Electron. Electr. Eng. 3(5), 101–104 (2015)

42

G. Harathi et al.

8. Singh, T., Kaur, M.: Smart dustbin for smart cities. Int. J. Comput. Sci. Inf. Technol. 7(2), 610–614 (2016) 9. Sharma, N., Singha, N., Dutta, T.: Smartbin implementation for smart cities. Int. J. Sci. Eng. Res. 6(9), 787–789 (2015) 10. Monika, K., Rao, N., Prapulla, S., Shobha, G.: Smart dustbin and efficient garbage monitoring system. Int. J. Eng. Sci. Comput. 6(6), 581–583 (2016) 11. Shubho, M.T.H., Hassan, M.T., Hossain, M.R., Neema, M.N.: Quantitative analysis of spatial pattern of dustbins and its pollution in the Dhaka city a GIS Based approach. Asian Trans. Eng. 3(4), 1–7 (2013)

Chapter 5

LabVIEW-Based Temperature Control Using Fuzzy Logic Controller C. H. Sai Krishna, S. Srinivasulu Raju, B. N. V. S. S. R. Dhanush, K. Harsha Vardhan, and K. R. M. V. Ganesh

Abstract This paperwork aims design, implementation of fuzzy logic controller for controlling of water in the tank using LabVIEW software. This system mainly contains the single tank, resistance temperature detector sensor, and myrio. The signal from the temperature sensor is transmitted to LabVIEW software via myrio interface, connected to the system. Fuzzy logic controller (FLC) algorithm is designed using knowledge-based inference engine, which can be used to replace the conventional controllers tuning parameters using analytical equations. The performing of fuzzy logic controller for controlling the temperature has been explored in detail. It provides an improved response and quickly tracks the setpoint. FLC functions well with the system involved uncertainties, noise at sensory signal. This developed FLC could also be used to measure several variables like temperature at industrial applications.

5.1 Introduction LabVIEW is a graphical user interface design environment that permits you to visualize each section of project, involving measurement data, configuration of hardware [1–4]. A process control system is a collection of electronic technology and appliances that maintains system stability and accuracy while preventing dangerous transition states in the process. In industries, the control system is widely used to own huge construction of continuous activities [5, 6]. Temperature control of water is a manner in which the variation in heat is observed and heat energy is controlled to accomplish the required temperature. It is a vital obligation in both residential and business applications because it offers a critical environment for combustion, chemical reactions, fermentation, drying, calcination, distillation, and other processes. Temperature management concerns can jeopardize crucial safety, quality, and productivity [7, 8]. After the industrial revolution it took a long time to find a fault in the system, C. H. Sai Krishna (B) · S. Srinivasulu Raju · B. N. V. S. S. R. Dhanush · K. Harsha Vardhan · K. R. M. V. Ganesh Department of EIE, VR Siddhartha Engineering College, Vijayawada, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_5

43

44

C. H. Sai Krishna et al.

Fig. 5.1 Front panel

especially with more complex process control systems [9]. As a result, these methods proved inflexible, and the need for a reliable and firm controller grew quickly, leading to the development of new technologies and software.

5.2 LabVIEW LabVIEW is a programme development application that uses a graphical programming language similar to a text-based language to construct programmes in a diagrammatic format hence it is called as an ‘universal programming system’ [10]. LabVIEW programme is additionally stated as virtual instruments. VI means the functions utilized in front panel for the purpose of operations and appearance that which imitate the particular instruments. The user interface and control of VI is identified as the front panel. The front panel is designed to look like a physical instrument’s panel. The virtual instrument is given instructions in the structure of a block graph. VI could also be used as a standalone programme or as a subprogram within another. The front panel is shown in Fig. 5.1. The purpose of the block diagram panel is to create a block diagram that connects the objects that send and/or receive data, perform specified functions, and govern execution flow. The LabVIEW block diagram is given in Fig. 5.2.

5.3 Temperature Controller To control a process temperature accurately with less help of operator, a temperature control structure depends on a controller that which receives the input from temperature sensor. The measured temperature is compared to required temperature, i.e., setpoint then delivers the output to a control section [11]. Considerable things in the

5 LabVIEW-Based Temperature Control Using Fuzzy Logic Controller

45

Fig. 5.2 Block diagram

selection of controller are, type of input sensor (RTD, thermocouple) and it’s range, control algorithm (PID), type of output essential (electromechanical relay, solid state relay, analogue output), and number and types of outputs required (heat, cold, alarm, and limit). When model for process is established, then control action is needed to maintain the process under steady state. Engineers and operators are required to control and monitor the process under the current system. This system also lacks real-time simulation, resulting in the complexity of text-based programming, as well as the difficulty of detecting and correcting problems. The use of LabVIEW software aids in the resolution of these issues and makes the process less complicated and time consuming. Cooling takes time in a typical temperature control operation. Real-time observing and control of many factors in an easier manner, with easy classification and correction of faults, is a requirement for today’s industry. The desire to speed up the cooling process emerges because the existing system’s cooling procedure is time consuming.

5.4 Fuzzy Systems Fuzzy logic controllers are commonly employed in manufacturing and product-based industries to track the expected reaction when the system’s analytical model is not exact. Fuzzy logic is a rule-based choice mechanism that is commonly employed in intelligent machines and production control. The membership of a user to the set is represented by the values one and zero, with one signifying absolute membership and 0 signifying no membership. Fuzzy rule for part participation, also known as a degree of membership, which can range from zero to one [12]. Fuzzy logic is mostly employed in the analysis and design of control systems for engineering issues that are extremely complex in nature and difficult to model. Expert knowledge and experience with the process will be necessary to frame the rule base that will be used to successfully track the robot vehicle system’s setpoints. Figure 5.3 depicts the basic architecture of fuzzy logic.

46

C. H. Sai Krishna et al.

Fig. 5.3 Overall configuration of fuzzy inference scheme

5.4.1 Linguistic Variables In other words, linguistic variables reflect the output and input variables of the system you want to regulate. Specify how several linguistic words, or classes of linguistics possible values, you wish to generate when creating an input variables to describe a digital output variable. Linguistic variables often have an unusual quantity of colloquial variables, with a medium element and symmetrical technical terms at both ends. Three to seven linguistic terms are usually enough to categorize the values of a linguistic variable in most cases. You could have two data language variables for a heater, desired and current temperature, and one and only output linguistic variable, heater setting, for example. There is a range of predicted values for each linguistic variable. The current temperature range, for example, could be 0–100°. The intended temperature range could be 50–80°. One input linguistic variable and one output linguistic variable are required for a fuzzy controller.

5.4.2 Membership Functions The membership tasks are mathematical functions that signify the degree of membership fun of linguistic variable quantity in terms and relate to linguistic terms. The scale of membership is a constant number among 0 and 1, with 0 equalling 0% membership and 1 equalling 100% membership. For instance, full membership (1) contained by the linguistic word hot at 100°, no membership (0) at 70° or less, and limited membership at every part of temperatures among 70 and 100° could be found for the linguistic variable current temperature.

5 LabVIEW-Based Temperature Control Using Fuzzy Logic Controller

47

Table 5.1 Rule base table E/CE

NB

NM

NS

ZE

PS

PM

PB

NB

NM

NS

Z

NS

NB

NB

NM

NM

NS

Z

PS

NS

NB

NM

NM

NS

NS

PS

PS

Z

NM

NM

NS

ZE

Z

PS

PM

PS

NM

NS

NS

PS

PS

PM

PM

PS

NS

NS

Z

PM

PS

PM

PB

PM

NS

Z

PS

PB

PM

PB

PB

PM

Z

PS

PS

5.5 Rule Base The fuzzy rules are outlined utilizing the knowledge base by contemplating the association among input and output of system. The rules are specified and described in Table 5.1, based on the expert knowledge about the system.

5.6 Fuzzy Controllers Fuzzy controllers can be used to regulate fuzzy systems. A quantitative system model to be controlled is required for most conventional control techniques. Many physical models, on the other hand, are hard or impossible to model numerically. Furthermore, many systems seem to be either unpredictable or too complex to be controlled using standard methods. One could use fuzzification to design a fuzzy controller that imitates a heuristics rule-of-thumb strategy if you can qualitatively characterize a control strategy.

5.6.1 Fuzzification In fuzzy set, input error covers seven triangular memberships as revealed in Fig. 5.4. Figure 5.4 shows the fuzzy set concerns to adjust in input error covers 7 triangular memberships. Figure 5.4 indicates the output of fuzzy set encompasses 7 triangular memberships.

48

C. H. Sai Krishna et al.

5.6.2 Implementing a Linguistic Control Strategy

NM

NS

Z

PS

PM

PB

0.8 0.6 0.4 0.2 0 -3

-2

-1

0

1

Input: Error (e)

2

3

1NB

NM

NS

Z

PS

PM

PB

0.8 0.6 0.4 0.2 0 -3

-2

-1

0

1

2

3

Degree of membership

1NB

Degree of membership

Degree of membership

Following the fuzzification of a fuzzy system’s input parameters, the fuzzy controller employs the interacting with the system linguistic terms as well as the rule base to identify the outputs linguistic variables’ consequent linguistic words. Let us say the present room temp is 50°, which correlates to a linguistic phrase for chilly with a membership degree of 0.4. Assume the intended temperature is 70° Fahrenheit, which is a linguistic phrase for mild with a degree of membership of 0.8. 1NB

NM

NS

Z

PS

PM

PB

-2

-1

0

1

2

3

0.8 0.6 0.4 0.2 0 -3

Input: Change in Error (ec)

Fig. 5.4 Membership funs of error, change in error, and output in fuzzy set

Output: U

5 LabVIEW-Based Temperature Control Using Fuzzy Logic Controller

49

Fig. 5.5 Block diagram of PID controller

5.6.3 Defuzzification Centroid technique is employed as a defuzzification method, which transforms the fuzzified value into the crisp value, and the fuzzy controller has been employed in LabVIEW environment.

5.7 Controller Design for Temperature Control The typical aim of control hypothesis is to get solutions from the controllers for the best feasible restorative action that results in framework dependability, which means the framework will retain the setpoint and not fluctuate around this one. Conventional controllers like P, PI, and PID controllers. PID controller holds proportional action, integral action, and derivative also and is indicated in Fig. 5.5.

5.8 Schematic Diagram The schematic diagram for the implementation of the fuzzy logic controller for the measuring the temperature is shown in Fig. 5.6. Fuzzy logic controller is used to measure and control the temperature of process tank.

5.9 Subsets for Inputs and Outputs 5.9.1 Input 1 (Current Temperature) Cold, moderate, and hot are taken as membership functions for input 1 shortly instead of PB, PM, PS, zero, NS, NM, and NB.

50

C. H. Sai Krishna et al.

Fig. 5.6 Schematic diagram

5.9.2 Input 2 (Desired Temperature) Cold, moderate, and hot are taken as membership functions for input 2 instead of positive big, positive medium, positive small and zero, negative small, negative medium, and negative big.

5.9.3 Output (Heater Temperature) Off, low ON, and high ON are taken as membership functions for Output instead of positive big, positive medium, positive small, zero, negative small, negative medium, and negative big.

5.10 The Rule Base Table See Table 5.1.

5.11 Proposed System The final stage in developing a fuzzy controller is to use the evaluation system to ensure that the defuzzified value is accurate. The LabVIEW system model is being used to evaluate the rule base of a fuzzy system by testing the connection among the

5 LabVIEW-Based Temperature Control Using Fuzzy Logic Controller

51

Fig. 5.7 Proposed system

parameters and output values. The quantities of the input parameters are manually entered in the test setup. The weights of the inputs are then calculated by the controller. The quantity of the output is calculated using the predecessor connector and resultant inference (Figs. 5.7 and 5.8).

5.12 Conclusion This article described a summary of PID controller, and design using Z-N method and layout of fuzzy logic control for temperature process system. The choosing of a controller is quite important. Controlling these plant characteristics will be a breeze after you have chosen the right controller. By using the fuzzy controller written in LabVIEW software, the proposed framework aids in the appropriate observing and regulation of water temperature in the tank. This system seeks to adjust the temperature difference between the measured and desired setpoints, resulting in effective temperature control. The demand of today’s residential or industrial purposes is for real-time checking and control of numerous factors in a more straightforward manner, with easy mistake detection and correction. Because the cooling in the present structure is time-consuming process, accelerating the cooling process is a pressing necessity, which the suggested system meets. The Ziegler Nichols adjusted

52

C. H. Sai Krishna et al.

Fig. 5.8 Results window

PID and the fuzzy logic control have been tested in LabVIEW, and the findings are presented here. In comparison with a PID, the FLC has no overshoot, zero steady state error, and a shorter settling time. In comparison with Ziegler Nichols adjusted PID control, the investigational findings show the suggested FLC with its easy model attitude and lesser rule base can deliver higher performance.

References 1. Sunay, A.S., Koçak, O., Kamberli, E., Koçum, C.: Design and construction of a LabVIEW based temperature controller with using fuzzy logic 2. Stephanopoulos, G.: Chemical Process Control. Pearson Education Publications (2009) 3. Vardhan, H., Bharadwaj, A.S., Raju, S.S., Archana, N.: Implementation of fuzzy PID controller and performance comparison with PID for position control of Dc motor. Int. J. Eng. Res. Dev. 7(9), 78–84 (2013) 4. Pushpaveni, T., Archana, N., Chandana, M.: Modelling and controlling of conical tank system using adaptive controllers and performance comparison with conventional PID. Int. J. Sci. Eng. Res. 4(5) (2013) 5. Fuzzy Logic Toolkit User Manual. www.ni.com 6. Clark, C.: LabVIEW Digital Signal Processing. McGraw-Hill Professional (2005) 7. Siddiqa, M.A., Ravi Kiran, T.K.S., Viswanath, M., Srinivasulu Raju, S.: Control of concentration in CSTR using DMC and conventional PID based on relay feedback method. Int. J. Eng. Sci. Technol. (IJEST) 7(4) (2013) 8. PID Controller, Wikipedia. [Online] Available: http://en.wikipedia.org/wiki/PID_controller

5 LabVIEW-Based Temperature Control Using Fuzzy Logic Controller

53

9. Darshan, T.S., Nagendra, B., Srinivasulu Raju, S.: Design of quadratic dynamic matrix control for driven pendulum system. Int. J. Electron. Commun. Eng. (IJECE) 5(3), 363–370 (2012). ISSN 0974-2166 10. Chattopadhyay, S., Roy, G., Panda, M.: Simple design of PID controller and tuning of its parameters using LabVIEW. Sens. Transducers J. 129(6), 69–85 (2011) 11. Kavitha, S., Ponmalar, S.J.: Fuzzy based control using lab view for temperature process. Int. J. Adv. Comput. Res. 2(4), 6 (2012) 12. Lee, C.C.: Fuzzy logic in control system: fuzzy logic controller—part I. IEEE Trans. Syst. Man Cybern. 20(2) (1990)

Chapter 6

A Survey Towards Implementing Smart Campus Anakhi Hazarika, K. D. K. Ajay, Nemani Subash, G. Srinivasa Yeshwanth, Lanka Raju, P. Kushal Swarup, S. K. Shireen Kasuar, and A. T. Antony

Abstract A smart campus is an environment with efficient technology and infrastructure to serve better facility and comfort. Smart campus includes smart classroom, libraries, cafeteria, parking place, transportation, etc., to improve the teaching process, students’ experience, and research. Also, the implementation of smart campus reduces overall operating costs. The Internet of Things (IoT) is considered as primary driving force in smart campus implementation. The IoT technologies with cloud/edge computing and cloud storage improve the space–time constraints of traditional campus environment. This paper reviews IoT-enabled smart campus applications and explores the probable research opportunities to be conducted in this area.

6.1 Introduction Smart campus technology is fetching its framework into reality by reducing human effort, enhancing digital technology that sources sustainable development, higher efficiency in operations, collaborations, and to transform people’s experience with digital technology. The concept of smart campus is similar to smart cities and its applications that create additional services and experiences to facilitate operational efficiency. Implementing a definite prototype model for smart cities is no more challenging with new technologies as it has defined objectives with finite ownership [1].

A. Hazarika Department of Electronics and Communication Engineering, Indian Institute of Information Technology Guwahati, Guwahati, India K. D. K. Ajay · N. Subash · G. Srinivasa Yeshwanth · L. Raju (B) · P. Kushal Swarup · S. K. Shireen Kasuar · A. T. Antony Department of Electronics and Communication Engineering, Malla Reddy College of Engineering and Technology, Hyderabad, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_6

55

56

A. Hazarika et al.

The area covered by a campus varies substantially depending on the university’s location, the financial endowment, and founding year. For example, we shall consider Malla Reddy College of Engineering and Technology (MRCET), Hyderabad, India. The college campus deployed on 160 acres with its arboriculture and horticulture, situated away from the hustle and bustle of the city, provides a serene atmosphere. Regardless of the initial surface area, it is expected that the campus will grow significantly as time goes on. Hence, institutions usually devise long-term sustainability plans to envision their growth in the future [2]. In an online learning control system (LMS), radio-frequency identification (RFID marker) tags are used on the phone or tablet to automatically transfer data to the group’s work environment. The notice is sent to other group members whose phones or tablets are logged into the LMS and automatically prompt at the appropriate time to release the information or access the website [3]. We need to go beyond content delivery to have the ability to immerse ourselves in immersed, customized, fully supported content that is genuinely available anytime and anywhere [4]. This paper aims to discuss the use of IoT technology in the smart campus development and modern IoT-enabled study halls where data collection is possible using E-learning application devices. It describes the importance of a smart classroom with IoT-enabled devices that download data on real-time activity. Also, this work helps to learn an overview of E-learning applications and IoT technology that integrate and utilize the services provided by the E-learning program for students by demonstrating its capabilities and performance.

6.2 Smart Campus—Concept and Benefits A campus is often described as an area that includes buildings for classrooms, student activity, a residential complex, a shopping center, libraries, parking, etc. Recently, campuses have been upgraded with several advanced technologies to provide quality services to the people (students, teaching, and non-teaching staff), academics, and administration in many aspects. The constant development in the campus environment reduces the operational cost and human efforts that offer a better facility and campus experience. The traditional campus has become a “smart campus” with several machine learning algorithms and IoT devices to analyze data and make effective decisions without human intervention. The concept of smart campuses revolutionizes the education system. It opens several research directions toward teaching and learning, data mining and analysis, energy efficiency in building management, water, and waste management, transportation, and security of the campus. Smart campuses offer the following advantages over traditional campuses [5]. (a) (b) (c) (d)

Save costs and time Automated maintenance Protection to campus environment Effective guidance to visitors in the campus

6 A Survey Towards Implementing Smart Campus

57

(e) Attendance monitoring on students and staff (f) Tracking the students inside the campus.

6.3 Survey of Smart Campus Applications

Author

Application

Technology/algorithm

Advancement

Limitations

Steven Hipwell

• Multi business protocol

• Service oriented architecture (SOA)

• Modified according to user interest • Provide students with best learning practices

• Records need to be monitored at regular intervals

• Analyze the student behavior/emotion using face recognition

• Cannot guide a fresher to his/her destination

• Record maintenance Weiyan Liang

Hanas Subakti et al

• Emotional analysis

• Face recognition

• Grasping power

• Analyzing the data using AI and ML

• Guidance system for freshers/new men

• Mysql data base

• We can go anywhere • Students can in the campus make use of it without any ones efficiently help

• Cyber-physical interaction (CPI) system, and indoor positioning (IP) system

• Can analyze and guarantee the security

• Augmented-reality (AR)

6.4 IoT Technology in Educational Sector The term smart campus is derived from the idea of smart cities that apply the principles of smart cities to campus operations [6]. Although many different work efforts have been ongoing, many existing reviews and activities focus on IoT technology. Specifically, not enough literature is available on IoT implementation in education applications [7]. Figure 6.1 demonstrates the IoT application in a few established needs for potential users, such as smart cities, intelligent power grid smart transport and mobility, smart homes, intelligent structures and infrastructure, intelligent industries and production, intelligent life, food and water security and safety; intelligent networks, and intelligent compasses. Technology has played a crucial role in the teaching of the university, which connects the three bodies: students, teachers, and

58

A. Hazarika et al.

Fig. 6.1 IOT applications as per review

administrative staff [8]. A significant impact in the field of education has changed the traditional teaching methods and developed and produced changes in the infrastructure of educational institutions, producing the idea of smart campuses [9]. The implementation of smart campuses should align with the institutional plan and the context in which the institution is established rather than related to the extensive use of communication technology [10, 11]. Several new avenues are open up on current challenges and future research trends in IoT-enabled campuses. First, the adoption of IoT on campus will lead to the use of thousands of sensors, actuators, and other devices, which can be a huge burden for manual configuration and maintenance. Indeed, it is necessary to thoroughly investigate new types of automatic configuration of its devices [10]. Data generated by the services such as infrastructure, management, education, and services serve as part of other programs, better than those intended for campus managers, in decision-making activities. Data mining from E-learning systems provides information about user’s interest, logs, and background knowledge. Blockchain technology may provide data security to user’s profile and preferences. Figure 6.2 illustrates the technologies used in the implementation of smart campuses or universities, such as cloud computing, the Internet of Things, big data, and artificial intelligence [12–14]. Machine learning algorithms analyze and learn the big data and deploy it on IoT devices. Cloud and/or edge computing is used for processing the data for smart devices. Some of the challenges faced while implementing a smart campus, including performance, mobility, reliability, security, and management, are major concerns [5, 15]. The term “smart university” (or) “smart campus” is still included in the textbook concept. Campuses or universities can be considered smart, because they make good use of available technologies to improve their performance and improve the quality of their students [16, 17].

6 A Survey Towards Implementing Smart Campus

59

Fig. 6.2 Technologies implemented in smart campus and smart universities

From the concept of smart university, an institute can be considered as smart, if it has the following [18–20]: . A special state of mind for the continuous improvement of education system and the ability to purpose that prioritizes formal and informal education. . A well-developed e-learning and m-learning system based on Web 2.0 tools, which provides students with extensive access to educational content from around the world, regardless of time, and place. . A well-developed power station within a developed framework. . Implementing the concept of a quick starting study in the university’s electronic framework. . Involvement of professionals working in curriculum development and effectiveness in the assessment process. . Smart reading; practical learning (visual laboratories); flexible teaching; shared reading; and learning games [21, 22].

6.5 Conclusion and Future Scope In this paper, the recent works on smart campus were investigated from different perceptions. A smart campus offers cost-effectiveness and high-quality services to the people associated with an institution. Deploying an IoT on several applications on campus is also highlighted in this paper. Integrating a smart campus often faces technical and financial challenges at the institution level. Privacy, standardization, and security are the primary technical issues concerned. Also, employing IoT devices within the university with limited resources creates a barrier in reality. Keeping the constraint in mind, the future work of smart campus should focus on on-campus services that did not get proper

60

A. Hazarika et al.

attention. Smart campus opens new avenues for the researchers, which provides optimized energy consumption, a learning environment, and an intelligent campus to the students within the campus.

References 1. Agarwal, P., GVV, R.K., Agarwal, P.: IoT based framework for smart campus: COVID-19 readiness. In: Fourth World Conference on Smart Trends in Systems, Security and Sustainability (WorldS4), pp. 539–542 (2020) 2. National Institute for Education Policy Research: Guide to the Creation of a Strategic Campus Master Plan. Available online: https://www.nier.go.jp/shisetsu/pdf/e-masterplan.pdf. Accessed on 31 Mar 2019 3. Veeramanickam, M.R.M., Mohanapriya, M.: IoT enabled futures smart campus with effective e-learning: I-campus. GSTF J. Eng. Technol. (JET) 3(4), 8–87 (2016) 4. Meyers, M.: Can the Internet of Things Make Education More Student-Focused? Deloitte Consulting LLP (2015). Retrieved from http://government2020.dupress.com/can-internet-thi ngs-make-education-student-focused/ 5. Abuarqoub, A., Abusaimeh, H., Hammoudeh, M., Uliyan, D, Abu-Hashem M.A., Murad, S., Al-Jarrah, M., Al-Fayez, F.: A survey on internet of things enabled smart campus applications. In: Proceedings of the International Conference on Future Networks and Distributed Systems, pp. 1–7 (2017) 6. Kar, A., Gupta, M.P.: How to make a Smart Campus—Smart Campus Programme in IIT Delhi (2015) 7. Gul, S., Asif, M., Ahmad, S., Yasir, M., Majid, M., Malik, M.S.A.: A survey on role of Internet of Things in education. Int. J. Comput. Sci. Netw. Secur. 17(5), 159–165 (2017) 8. Mohamed, S.E.: Experimental evaluation of Internet of Things in the educational environment. Int. J. Eng. 7(3), 50–60 (2017) 9. Pandey, R., Verma, M.: Current emerging trends in IoT: a survey and future prospects. 8(ii), 339–344 10. Arsan, T.: Smart systems: from design to implementation of embedded Smart Systems. In: HONET-ICT, pp. 59–64 (2016) 11. Zhan, S.: The reconstruction strategy “Internet+” from the perspective of education (2017) 12. Broadband Internet Technical Advisory Group (BITAG): Internet of Things (IoT) security and privacy recommendations. Technical report (2016) 13. Flauzac, O., Gonzalez, C., Nolot, F.: New security architecture for IoT network. Procedia Comput. Sci. 52(1), 1028–1033 (2015) 14. Sánchez-Torres, B., Rodríguez-Rodríguez, J.A., Rico-Bautista, D.W., Guerrero, C.D.: Smart campus: trends in cyber security and future development. Rev. Fac. Ing. 27(47) (2018) 15. Li, S., Da Xu, L., Zhao, S.: The Internet of Things: a survey. Inf. Syst. Front. 17(2), 243–259 (2015) 16. Silva, B.N., Khan, M., Han, K.: Towards sustainable smart cities: a review of trends, architectures, components, and open challenges in smart cities. Sustain. Cities Soc. 38, 697–713 (2018) 17. Bagheri, M., Movahed, S.H.: The effect of the Internet of Things (IoT) on education business model. In: 12th International Conference on Signal-Image Technology and Internet Based Systems (SITIS), pp. 435–441 (2016) 18. Rehman, A.Z.A., Shaikh, Z.A.: Building a smart university using RFID technology. In: International Conference on Computer Science and Software Engineering, pp. 641–644 (2008) 19. Gierej, S.: The framework of business model in the context of industrial Internet of Things. Procedia Eng. 182, 206–212 (2017)

6 A Survey Towards Implementing Smart Campus

61

20. Silva, J.D.C., Rodrigues, J.D., Mario: IoT network management: content and analysis. SBrT (2017) 21. Bueno-Delgado, M.V., Pavón-Marino, P., De-Gea-García, A., Dolón-García, A.: The smart university experience: an NFC-based ubiquitous environment. In: Sixth International Conference on Innovative Mobile and Internet Services in Ubiquitous Computing, pp. 799–804 (2012) 22. Semenova, N.V., Svyatkina, E.A., Pismak, T.G., Polezhaeva, Z.Y.: The realities of smart education in the contemporary Russian Universities. In: Proceedings of the International Conference on Electronic Governance and Open Society: Challenges in Eurasia, pp. 48–52 (2017)

Chapter 7

Design and Analysis of Stepped Impedance Feed Elliptical Patch Antenna P. Sree Lakshmi, A. Deepak, Suresh Kumar Muthuvel, and Ch. Amarnatha Sarma

Abstract Nowadays, the requirement for wireless communication increased rapidly with the demand for applications. The antenna usually operates in single band operation, but it is more useful when it operated in dual, multi-bands or wideband. The microstrip patch antenna is more popular because of compactness and easy fabrication. In this work, an elliptical patch antenna is designed on the 1.6 mm thickness FR4 substrate with εr = 4.2 and 0.02 loss tangent. In this, stepped impedance type feed is designed and analyzed. The size of the antenna is 40 mm × 30 mm. The antenna simulation results are offer dual bands of operation with 4 GHz bandwidth each. One is from 3 to 7.8 GHz, and the other is from 12 to 16 GHz, respectively, and field strengths of − 3.6 and – 6 dBi, respectively, and maximum peak gain is − 1.3 dB. Measured return loss (S11) has good operation band from 3.5 to 10.5 GHz. The band operation shows the antenna cover the application in C and X bands applications efficiently.

7.1 Introduction With the huge requirements of society, these wireless communication systems growths are not having limits. The major challenges are compactness, efficiency P. Sree Lakshmi Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences, Saveetha University, Chennai, Tamil Nadu, India A. Deepak · S. K. Muthuvel Deapartment of Nano Elelctronics Materials and Sensors, Saveetha School of Engineering, Saveetha Institute of Medical and Technical Sciences, Saveetha University, Chennai, Tamil Nadu, India e-mail: [email protected] Ch. Amarnatha Sarma (B) Faculty, Department of Electronics and Communication Engineering, RSR Engineering College, SPSR Nellore, Andhra Pradesh, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_7

63

64

P. Sree Lakshmi et al.

and power handling capabilities. The power handling capabilities can be improved by proper circuit design along with the antenna. But the performance of handling a wide band of allocation depends on the antenna design. The elliptical and circular shape antennas with lesser size than rectangular patch can show better performance. In rectangular and circular structures there is no freedom in dimensions and feeding places on the patch. In ellipse, there is some freedom for the designer to change the geometry of the elliptical patch so the size can be reduced [1]. The band of frequencies from 3.5 to 10.6 GHz is approved as UWB by FCC in 2002. Various commercial applications come under this frequency band. So the design of the antenna in the frequency range of 3–11 GHz this band is a profitable challenge [2], because most applications are in this and various methods such as ground etching and providing slots in the patch are used to get this wideband and ultra-wideband antenna structures, but the design complexity is more. The microstrip patch with elliptical structure is proposed in [3], the design is very simple, but its return loss curve is not good below 10 GHz. The antenna proposed in [4] is a multi-band operative and elliptical shape. But all bands are narrow band. In [5–8] all are modified structures of patch and each structure can give motivation for improvements in antenna parameters. Removing a part of ground behind patch area can convert the monopole antenna to dipole antenna. Inset feed is shown an improvement in frequency response in some designs. Changing the length and width of the inset feed the wideband characteristics also change [9–14]. Based on the above work reports, an elliptical patch antenna is proposing here that offer dual band operation for C band and K band applications. In the design, a stepped impedance feeding is used. The simulation is done using commercial EM simulator HFSS and printed on FR4 substrate. Rest of the paper consist three sections. Section 7.2 is design process given, Sect. 7.3 is simulation and comparison with measured results, and Sect. 7.4 is conclusion of the work.

7.2 Design of Stepped Impedance Feed Elliptical Patch Antenna The elliptical-shaped patch antennas are designed using analytical and genetic algorithms. In this paper, the area of the elliptical patch and the dimensions are calculated based on analytical method using the following metrics [4, 11]: Elliptical Patcharea = 3.14(Ma m b )

(7.1)

Here, M a and mb are effective lengths of semi-major and minor axes, respectively. They are calculated using / R=

X2

[ ] X h 2h X ln + 1.41εeff + 1.77 + [0.268εeff + 1.65] + π εeff 2h X

(7.2)

7 Design and Analysis of Stepped Impedance Feed Elliptical Patch Antenna

65

In above equation, R = M a , when X = a; R = ma , when X = b εr eff

εr + 1 εr − 1 + = 2 2

/[

h 1 + 12 w

]−1 (7.3)

h is the height of the substrate; w = width of substrate; εr is dielectric constant of the substrate. Using above equations, the elliptical patch is having major axis of length M = 2M a and minor axis of length m = 2mb shown in Fig. 7.1. The antenna is fabricated on FR4 substrate having length (L s ) and width (W s ) of 40 mm and 30 mm with substrate height of 1.6 mm. Its major axis (M) = 24 mm, minor axis (m) = 12 mm, fed line length (L f ) and width (W f ) are 22 mm and 1.6 mm, respectively. Its simulation results are shown in Fig. 7.2. The simulation results shown in Fig. 7.2a are return loss (S11), and Fig. 7.2b is electric field strengths. The highlighted region in S11 shows pass band which is the radiation band and the electric filed plot is the radiation pattern in E plane (black line) and H-plane (red line). The S11 shown in Fig. 7.2a is narrow band, and the radiation plot is monopole radiation pattern. But antenna with wide band and dipole radiation Fig. 7.1 Elliptical patch antenna with full ground. a Top view and b bottom view

Fig. 7.2 Elliptical patch antenna with full ground. a Return loss (S11) and b radiation pattern

66

P. Sree Lakshmi et al.

Fig. 7.3 Elliptical patch antenna with etched ground. a Top view and b bottom view

Fig. 7.4 Elliptical patch antenna with partial ground. a Return loss (S11) and b radiation pattern

characteristics is recommended. To improve results, some part of ground is removed till patch are elliptical patch with partial ground below feed line. Ground under patch is removed so it left with length (L f ) of 22 mm ground as shown in Fig. 7.3. The resultant return loss (S11) and the radiation plot are shown in Fig. 7.4. Above results show that the, by removing ground below the patch area is significantly improved results in both return loss and radiation pattern. But, with the knowledge from literature, the return loss (S11) can be improved more by changing the microstrip feed line. In this work, a new stepped impedance structure is provided in feed line. This stepped impedance structure basically a low pass filter that offers good return loss. By adding patch at one port, it may get radiate with improved return loss for both bands. The new feed line is cascading of 5 resonators of lengths L1 = 6 mm, L2 = 3 mm, L3 = 5 mm, L4 = 4 mm, L5 = 5 mm, widths are W f and W 1.6 mm and 1 mm, respectively. The construction shown in Fig. 7.5a and its simulation results are shown in same Fig. 7.5b–e.

7 Design and Analysis of Stepped Impedance Feed Elliptical Patch Antenna

67

Fig. 7.5 Stepped impedance feed elliptical patch antenna with partial ground: a top view and bottom, b return loss (S11), c peak gain in dB, d radiation pattern at 5.5 GHz, e radiation pattern at 14 GHz

7.3 Results and Discussion From simulation results shown in Figs. 7.2 and 7.4, it is observed that the antenna offers dipole radiation pattern when ground is partially removed. The return loss is improved, and the narrow band operation became wide band. Radiation pattern also changed from single direction to bidirectional pattern. In Fig. 7.4, the return loss in the second band is poor. One way to improve return loss is changing the feed path. In this, stepped impedance feed is applied. The improvement in results is shown in Fig. 7.5, and it offers dual band operation. The final model is able to radiate in at from 3 to 7.8 GHz with maximum radiation at 5.5 GHz and wide band above 12 GHz with maximum radiation at 14 GHz. The radiated field strengths are shown in Fig. 7.5c and d. The consolidated results are shown in Fig. 7.6. In this figure, it shown that the elliptical patch with partial ground and steeped impedance feed is having very low return loss in the bands, 3–7.8 GHz and above 12 GHz. Figure 7.7 shows fabricated antenna and measurement setup and the comparison of measured and simulated results of final model is shown in Fig. 7.8. The measured results are good at single band rather than dual band. This single band has a band width of 7.5 GHz from 3.5 to 11 GHz.

68

P. Sree Lakshmi et al.

Fig. 7.6 Variation in return loss (S11) with inset feed length (L i )

(a)

(b)

(c)

Fig. 7.7 Fabricated antenna. a Top view, b bottom view, c S11 measurement setup

7.4 Conclusion In this work, the results shows that with removal of ground below patch will get the dipole radiation pattern. Changing feeding line will improve the antenna parameters also shown. This measure and simulated results are almost same with acceptable deviations. The measured results of proposed model showing good return loss from 3.5 to 11 GHz, i.e., almost ultra wide band operation and poor return loss in other band form 13.5 GHz. Even the radiation characteristics shown in Fig. 7.5e also support that it has poor operation beyond 10 GHz. So it is recommended that the antenna could give better performance in single band rather than dual band.

7 Design and Analysis of Stepped Impedance Feed Elliptical Patch Antenna

69

Fig. 7.8 Comparison of simulated and measured return loss (S11) results and radiation pattern

References 1. Sharma, M.M., Yadav, S., Kumar, A., Ranga, Y., Bhatnagar, D.: Compact elliptical microstrip patch antenna with slotted ground for Ku-band applications. In: IEEE Applied Electromagnetics Conference (AEMC), Kolkata, pp. 1–3 (2011) 2. Awad, N.M., Abdelazeez, M.K.: Multislot microstrip antenna for ultra-wide band applications. J. King Saud Univ. —Eng. Sci. 30(1), 38–45 (2018) 3. Rama Krishna, C., Ganesh, N., Prasad, D.D.: Design of elliptical shaped micro-strip patch antenna for Ka Ban. Int. J. Res. Anal. Rev. 5(3), 961–965 (2018) 4. Jose, J.V., Shobha Rekh, A., Jose, M.J.: Design techniques for elliptical micro-strip patch antenna and their effects on antenna performance. Int. J. Innovative Technol. Explor. Eng. (IJITEE) 8(12), 2317–2326 (2019). ISSN: 2278-3075 5. Gupta, M., Mathur, V.: Multiband: multiple elliptical microstrip patch antenna with circular polarization. Wirel. Pers. Commun. 102, 355–368 (2018) 6. Jung, H., Seo, C.: Analysis of elliptical microstrip patch antenna considering attachment mode. IEEE Trans. Antennas Propag. 50(6), 888–890 (2002) 7. Sharma, V., Saxena, V.K., Sharma, K.B., Bhatnagar, D.: Radiation performance of an elliptical patch antenna with three orthogonal sector slots. Rom. J. Inf. Sci. Technol. 14(2), 123–130 (2011) 8. Jose, J.V., Paulson, A.S.R., Jose, M.J.: Double-elliptical micro-strip patch antenna for higher design flexibility and miniaturization. Int. J. Eng. Adv. Technol. 9(1), 6970–6976 (2019)

70

P. Sree Lakshmi et al.

9. Samarthay, V., Pundir, S., Lal, B.: Designing and optimization of inset fed rectangular microstrip patch antenna (RMPA) for varying inset gap and inset length. Int. J. Electron. Electr. Eng. 7(9), 1007–1013 (2014) 10. Shankar, S., Chaurasiya, H.: Inset feed microstrip patch antenna. In: International Conference On Computer, Communication And Control (IC4), Indore, pp. 1–3 (2015) 11. Sarma, C.A., Inthiyaz, S., Madhav, B.T.P.: Effect of ground etching, inset feed and substrate height on elliptically shaped patch antenna. Int. J. Emerg. Trends Eng. Res. 8(7), 3145–3149 (2020) 12. Hu, Y., Jackson, D.R., Williams, J.T., Long, S.A.: A design approach for inset-fed rectangular microstrip antennas. In: IEEE Antennas and Propagation Society International Symposium, Albuquerque, NM, pp. 1491–1494 (2006) 13. Josan, S.K., Sohal, J.S., Dhaliwal, B.S.: Design of elliptical microstrip patch antenna using genetic algorithms. In: IEEE International Conference on Communication Systems (ICCS), Singapore, pp. 140–143 (2012) 14. Zhang, Z., Yang, S., Liu, M., Deng, S., Li, L.: Design of an UWB microstrip antenna with DGS based on genetic algorithm. In: 21st International Conference on Advanced Communication Technology (ICACT), Pyeongchang Kwangwoon_Do, Korea (South), pp. 228–232 (2019)

Chapter 8

Wavelet-Based Colon Polyp Detection Using Support Vector Machine Classifier B. Jyothi, M. Sucharitha, and Anitha Patibandla

Abstract One of the leading cancer-related causes of mortality, particularly among men, is colorectal cancer and early detection is important. Unusual growths called colonic polyps have the potential to develop into colon cancer. One of the major contributors to colorectal cancer are polyps, which a colonoscopy can detect early on and treat successfully. Elevated sensitivities and small-scale faulty positive rates are regularly attained since detection of small size or bigger polyps, inferior sensitivities and greater deceitful affirmative rates arise where the purpose of CAD is to categorize moderate dimension polyps. Colorectal cancer incidence has been proven to be decreased by the recognition and elimination of colonic polyps, especially bigger sized. It takes a lot of time and effort to diagnose colorectal disease manually. Because polyps can vary in size and shape, diagnosing them can be difficult. In this study, we introduce a technique for segmenting and classifying colorectal diseases automatically. We suggested a Wavelet-based polyp segmentation technique. Two tactics improve the method’s performance. First, after applying the wavelet transformation, we receive the clusters. After completing the morphological operation, we then perform an efficient grouping via a SVM classifier in the test phase. When compared to earlier colonoscopy segmentation techniques, our suggested method produces results that are more accurate, by utilizing the CVC-ColonDB database.

8.1 Introduction Although colon cancer is a preventable cancer, it is a cause of cancer mortality in the US [1]. The best method of preventing colon cancer is early discovery and elimination of colon polyps, which are tiny extensions in the colon lining. Computer B. Jyothi · A. Patibandla Department of ECE, Malla Reddy College of Engineering and Technology, Dhulapally, Secunderabad, India M. Sucharitha (B) School of Electronics Engineering, VIT-AP University, Amaravati, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_8

71

72

B. Jyothi et al.

assisted detection (CAD) systems [2] are proposed as a method to help the doctor read these tests. Contemporary CT colonography (CTC) CAD systems incline to identify erroneous positives after detecting tiny 6–9 mm polyps, but have good sensitivity and truncated deceitful affirmative rates for polyps 1 cm or bigger in thickness [3]. Adults aged 65–74 in the United States are most likely to develop colorectal cancer. Due to increased screening and improvements in some risk factors, the number of novel cases of colorectal cancer in persons with 50 years or elder is declining [4]. The goal of this paper is to create a postprocessor which employs wavelet examination of the endoluminal prognosis imageries to significantly lower the quantity of incorrect predictions from a CTC CAD structure for the 6–9 mm polyps based on the clinical interpretation process of 3D virtual flythrough reading. The following sections make up the paper’s structure. Section 8.2 delivers a brief explanation of the effort associated with this paper. Section 8.3 describes wavelet, curvelet and wave atom transform principles. Section 8.4 discusses observations, and the final section draws a conclusion.

8.2 Related Work One of the most prevalent and easily curable types of cancer is colon cancer. The best strategy for therapy is early diagnosis of colon polyps. Although the current procedures for colonoscopy, which are not ideal, virtual colonoscopy is a promising new option. This method uses CT scans to identify polyps, often by a CAD detection system. We have been creating a CTC CAD application that uses polyp curvature to identify them [5]. Feature extraction uses wavelets with extensive uses in the technical domain, specifically pattern recognition chores. Discriminant analysis serves as an effective tool for wavelet decomposition of an image with output coefficients. Wavelet coefficients are employed in feature extraction methodologies. An efficient method for revolution and measure unvaried surface grouping via log-polar wavelet signatures was proposed by Pun and Lee [6]. Output wavelet coefficients achieved are revolution and gage unvariable. Khouzani et al. [7] extract features using multi-wavelet and wavelet packet transforms. An energy and entropy feature on each sub-band is obtained by the wavelet decomposition. The image processing works shows a number of de-noising methods built on Partial Differential Equations (PDEs) [8], comprising some of them focused on MR Images [9]. Although, these methods have the benefit of simplicity, removal of stair case outcome that follows with TV-norm filter. Alternative style to image refurbishment is nonparametric analytical approaches. For illustration, in [10] suggested an unsubstantiated info-theoretic robust filter, namely UINTA that depends on nonparametric MRF prototypes resulting from the degraded imageries. UINTA reinstates pictures by simplifying the mean-shift technique [11] to integrate vicinity info. Though the preference still rests in the

8 Wavelet-Based Colon Polyp Detection Using Support Vector Machine …

73

scaling factors, it is not signal-reliant and can hence be simply detached [12, 13]. Equally, de-noising methods established on anisotropic diffusion are recommended [14]. The striving with wavelet or anisotropic diffusion procedures is the jeopardy of over-smoothing fine specifics principally in small SNR pictures [15]. Lately, various prevalent de-noising processes proposed are built on wavelet thresholding. These methodologies try to isolate noteworthy features/signals from noise in the frequency realm and concurrently reserve them but eliminate noise. Manry et al. [16] suggests an effectual feature selection procedure for the universal regression problem utilizing a piecewise linear orthonormal least squares technique. Proposed procedure is reckoning very effectual since only one data permit is compulsory. Wavelet-based method is of significance and used lately to ascertain counterfeits in treasured images. A novel wavelet feature extraction technique to discriminate actual polyps from deceitful positives, where we selected 52 detection images for discriminant investigation. Numerous features were added and an efficient process for feature extraction is employed and the suggested technique on all findings imageries formed by the CTC CAD program applied on 44 patient’s dataset.

8.3 Proposed System In order to detect the polyps, wavelet technique and morphological operation is applied on the acquired CT images. Accurate polyp is obtained by using the SVM method. Source image is of DICOM image format and is converted into 2-dimension image format. Unwanted information from the image is eliminated by preprocessing the image. Output pixel value is obtained by calculating the vicinity pixels central value and is thus capable of eliminating the outliers deprived of decreasing the intelligence of the image. De-noising of image is performed using wavelet, curvelet and wave atom approaches.

8.3.1 Wavelet Wavelet sources are sources of encapsulated function places, which can be applied to investigate signals at several gauges. The time and frequency information are contained in wavelet coefficients as and the essential tasks differ in place and gauge. A signal is converted to its wavelet illustration [15] by applying fast wavelet transform (FWT). In a single-level FWT, a signal is fragmented into an approximation and a specific portion. In a multi-lateral FWT, every successive is split into an approximation and specific portion. For 2-D images, all successive portions are divided into an approximation and three detailed stations as horizontal, vertical and diagonal focused on particulars, correspondingly. The inverse FWT (IFWT) recreates

74

B. Jyothi et al.

every successive from approximation and specific stations. If the wavelet base tasks lag compact maintenance, the FWT is computed most competently in the frequency domain. X (t) = S(t) + N (t)

(8.1)

If the experimental statistics comprises of the accurate signal S(t) with added noise N(t) as tasks in time t to be sampled. Let W (.) and W − 1(.) signify the forward and inverse wavelet transform operators. Let D(., λ) signify the de-noising operative with soft threshold λ. To wavelet shrinkage de-noise X(t) in order to recuperate S (t) as an estimate of S(t). 

Take the forward transform, Y = W (X )

(8.2)

Z = D(Y ; _)

(8.3)



S = W − 1(Z )

(8.4)

summarize the procedure.

8.3.2 Curvelet The curvelet transform, similar to the wavelet transform, is a multiscale transform containing frame elements ordered by range and positional constraints. In contrast to the wavelet transform, indicator constraints and the curvelet pyramid [17] comprises essentials having an exact gradation of steering relevance. The foundations follow an exceptional mounting rule, where the span of the provision of a frame foundation and the thickness of the sustenance are connected by the relative width = length2 . Curvelets are exciting since they proficiently report highly vital hitches where wavelets are outlying since perfect. Curvelets deliver extraordinary sparse depictions of items, which exhibit curvepunctuated evenness excluding incoherence beside a common curve with constrained curvature. Such depictions are almost as light as if the entity is not remarkable and is lighter compared to the wavelet disintegration of the entity. The incident finds rapid uses in estimation postulate as well as in algebraic approximation. In estimate theory, assume f m be the m-term curvelet estimation (conforming to the m leading coefficients in the curvelet sequence) to an entity f (x 1 , x 2 ) ε L 2 (R2 ). The augmented scantiness expresses that if the entity f is remarkable laterally a standard flat C 2 curve but smooth, the estimation inaccuracy conforms  f − f m 2L 2 ≤ C.(log m)3 .m −2

(8.5)

8 Wavelet-Based Colon Polyp Detection Using Support Vector Machine …

75

and is optimum in the perception that no alternative illustration can produce a minor asymptotic inaccuracy with the similar quantity of terms. The inference in data is that one can retrieve such entities via noisy statistics through modest curvelet decline and attain an MSE of scale superior comparable to that might be accomplished by other classical techniques.

8.3.3 Wave Atom Demanet and Ying [18] familiarized wave atoms, which are variant of 2-D wavelet packs and follow the parabolic mounting of curvelets wavelength = (diameter)2 . Oscillatory tasks or positional surfaces (e.g., fingerprint, seismic contour, engineering outsides) have an expressively lighter increase in wave atoms comparable to stationary customary illustrations identical to Gabor filters, wavelets and curvelets. Wave atoms require the capability to adjust to random local orders of a design, and to sparingly characterize anisotropic forms affiliated with the axes. In contrast to curvelets, wave atoms, acquire the consistency of the shape laterally the alternations and the outline crossways the fluctuations. Discrete wavelet transform (DWT) acquires positional and frequency data which separate frequency detail. It is a multiresolution decomposition and can split the imagery into four sub-imageries at a quartier dimensions, i.e., small frequency estimated sub-imagery, and three horizontal, vertical and diagonal way higher frequency specifics sub-imageries. In our proposed method, Harr and Daubechies wavelets are used for wavelet transformation (Fig. 8.1).

CT Image

Preprocessing

Morphological Operation

Clustering

Support Vector Machine

Output

Fig. 8.1 Proposed method

Color Conversion

Wavelet transform

76

B. Jyothi et al.

8.3.4 Experiments and Results This segment discusses a comprehensive analysis of the projected segmentation process. It is evaluated with simulated imageries and correlates the metrics of the projected technique with K-means clustering algorithm. Input images See Figs. 8.2, 8.3, 8.4, 8.5 and 8.6.

Fig. 8.2 a Image 1, b image 2

Fig. 8.3 Kirsch’s template for image 1

8 Wavelet-Based Colon Polyp Detection Using Support Vector Machine …

77

Fig. 8.4 Segmented output for image 1

Fig. 8.5 Kirsch’s template for image 2

Figure shows the segmented output obtained for the proposed method using two images. Segmentation separates the polyps region from the non-polyp region and is useful for the further diagnosis of disease. Table 8.1 displays the TP, FP, TN and FN attained from the SVM classifier and are compared with the existing KNN classifier. Table 8.2 represents the accuracy value, F1 score and MCC for polyp detection and our proposed method achieved an better accuracy of 97.5%.

78

B. Jyothi et al.

Fig. 8.6 Segmented output for image 2

Table 8.1 Classifier output

Table 8.2 Performance parameters for colon detection

Parameter

K-nearest network

SVM network

TP

18

19

FP

2

1

TN

20

20

FN

0

0

Parameter

K-nearest network

SVM network

ACC %

95

97.5

F1%

94.74

97.44

MCC

0.9045

0.9514

8.4 Conclusion Our proposed method is to detect colorectal cancer at an initial period to reduce its threat, which frequently causes mortality. A novel system is anticipated for the denoising of MRI scans by means of wave atom contraction. De-noising method was used to remove the noise and then wavelet features are extracted from the preprocessed image. Solitary DWT was enforced via Haar wavelets. Morphological operations are performed on the clustered image and finally by using the classifier classified as belongs to polyps or non-polyps. The SVM outcomes are outstanding, with little fault for the categorization scheme, while the outcomes of the KNN were satisfactory. An automatic structure is created for the classification and disclosure of colorectal cancer with virtuous consequences, evading the drawbacks of classical techniques. In future work, deep learning network can be adapted, and diverse classifiers could be selected to enhance the outcomes.

8 Wavelet-Based Colon Polyp Detection Using Support Vector Machine …

79

References 1. Siegel, R.L., Miller, K.D., Goding Sauer, A., et al.: Colorectal cancer statistics. CA: A Cancer J. Clin. 70(3), 145–164 (2020) 2. Virostko, J., Capasso, A., Yankeelov, T.E., Goodgame, B.: Recent trends in the age at diagnosis of colorectal cancer in the US National Cancer Data Base, 2004–2015. Cancer 125(21), 3828– 3835 (2019) 3. Ansa, B.E., Coughlin, S.S., Alema-Mensah, E., Smith, S.A.: Evaluation of colorectal cancer incidence trends in the United States (2000–2014). J. Clin. Med. 7(2), 22 (2018) 4. Kehm, R.D., Yang, W., Tehranifar, P., Terry, M.B.: 40 years of change in age- and stage-specific cancer incidence rates in US women and men. JNCI Cancer Spectr. 3(3) (2019) 5. Summers, R.M., Beaulieu, C.F., Pusanik, L.M., Malley, J.D., Jeffrey, R.B., Glazer, D.I., Napel, S.: Radiology 216(1), 284–290 (2000) 6. Pun, C.M., Lee, M.C.: Log-polar wavelet energy signatures for rotation and scale invariant texture classification. IEEE. Tran. Pat. Ana. Mach. Intel. 25, 590–603 (2003) 7. Khouzani, K.J., Siadat, M.R., Zadeh, H.S., Elisevich, K.: Texture analysis of hippocampus for epilepsy. In: Amini, A.A. (ed.) Proceeding of SPIE Medical Imaging, vol. 5031, pp. 279–288 (2003) 8. Tai, X., Lie, K., Chan, T., Osher, S. (eds.): Image Processing Based on Partial Differential Equations. Springer, New York (2005) 9. Gerig, G., Kubler, O., Kikinis, R., Jolesz, F.A.: Nonlinear anisotropic filtering of MRI data. IEEE Trans. Med. Imag. 11(2), 221–232 (1992) 10. Awate, S.P., Whitaker, R.T.: Higher-order image statistics for unsupervised, informationtheoretic, adaptive, image filtering. In: Proceedings of the IEEE International Conference on Computer Vision Pattern Recognition, vol. 2, pp. 44–51 (2005) 11. Comaniciu, D., Meer, P.: Mean shift: a robust approach toward feature space analysis. IEEE Trans. Pattern Anal. Mach. Intell. 24(5), 603–619 (2002) 12. Nowak, R.D.: Wavelet-based Rician noise removal for magnetic resonance imaging. IEEE Trans. Image Process. 8(10), 1408–1419 (1999) 13. Pizurica, A., Wink, A.M., Vansteenkiste, E., Philips, W., Roerdink, J.B.T.M.: A review of wavelet denoising in MRI and ultrasound brain imaging. Current Med. Imag. Rev. 2(2), 247– 260 (2006) 14. Perona, P., Malik, J.: Scale-space and edge detection using anisotropic diffusion. IEEE Trans. Pattern Anal. Mach. Intell. 12(7), 629–639 (1990) 15. Samsonov, A.A., Johnson, C.R.: Noise-adaptive nonlinear diffusion filtering of MR images with spatially varying noise levels. Magn. Reson. Med. 52, 798–806 (2004) 16. Li, J., Manry, M.T., Narasimha, P.L., Yu, C.: Feature selection using a piecewise linear network. IEEE Tran. Neural Network 17. Do, M.N., Vetterli, M.: Framing pyramids. IEEE Trans. Signal Proc. 2329–2342 (2003) 18. Demanet, L., Ying, L.: Wave atoms and sparsity of oscillatory patterns. Appl. Comput. Harmon. Anal. 23(3), 368–387 (2007)

Chapter 9

Probable Deviation Outlier-Based Classification of Obesity with Eating Habits and Physical Condition M. Shyamala Devi, P. S. Ramesh, Aparna Joshi, K. Maithili, and A. Prem Chand Abstract Obesity is harmful because it would be associated to poor mental health and lower living standards. Obesity has also been linked to several of the world’s most common causes of death, including diabetes, heart disease, stroke and certain types of cancer. Obesity is a serious health problem that can be influenced by a multitude of factors, including genetics and behavior. Because the original cause of obesity is unknown, determining the levels of obesity remains a difficult task for doctors and researchers. With this overview, this project uses machine learning algorithms to identify obesity levels based on eating habits and physical condition. The obesity dataset from the KAGGLE database repository is used for implementation for the classification of obesity based on eating habits and physical condition. The obesity dataset contains 17 features and 2111 patient details and it is preprocessed with encoding and missing values. To analyze the performance metrics, the original dataset is used with all classifiers, both with and without feature scaling. The exploratory data analysis is done to figure out how the target obesity variable is distributed. The dataset is analyzed with the scatterplot to detect the features that are having outliers. From the implementation, it is observed that the features age and weight have outliers. The dataset is applied with outlier detection methods like interquartile range and standard deviation to balance the outliers in age and weight features. The outlier is removed to form the following dataset as age outlier removed using STD, age outlier removed using IQR, weight outlier removed using STD, age and weight outlier removed using STD, age and weight outlier removed using IQR. The above outlier removed datasets are applied with various classifiers to predict the obesity levels before and after feature scaling to analyze the performances indices precision, recall, f score and accuracy. Experimental results shows that the Decision Tree classifier shows accuracy of 94% before removing outliers on age and weight M. Shyamala Devi (B) · P. S. Ramesh · K. Maithili · A. Prem Chand Computer Science and Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Chennai, Tamilnadu, India e-mail: [email protected] A. Joshi Department of Information Technology, Army Institute of Technology, Pune, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_9

81

82

M. Shyamala Devi et al.

feature. The same Decision Tree classifier shows the accuracy of 96% after removing outliers on age and weight feature.

9.1 Introduction The correct interpretation of health data promotes early illness identification, patient treatment and outreach programs, in view of the increasing rise of machine learning in the healthcare and medical professions. Whenever the input and provision of health data is untapped and inadequate, nevertheless, the precision of illness diagnoses is lowered. Obesity is an often misunderstood but serious health problem that can lead to a variety of potentially fatal cardiovascular illnesses if left untreated. Many people ignore this problem, not recognizing that by making simple changes to their healthcare, they can prevent many additional ailments. Numerous machine learning research have been presented to construct obesity forecasting model or to identify major factors of obesity in required to formulate prevention methods. Prevalence of overweight and obesity for both children and young adults is one of the all high, despite a huge and varied body of literature on obesity prediction models. In order to design effective measures to minimize adolescent and adult obesity, a thorough understanding and critical examination of existing machine learning models is required. The paper is organized in which literature review is discussed in Sect. 9.2 followed by the paper contributions in Sect. 9.3. The implementation setup and results are discussed in Sect. 9.4 and concluded in Sect. 9.5.

9.2 Literature Review India’s increasingly profession lifestyle leads to abnormal biological configurations, especially to the younger millennium, who recommend games online to having played outdoor. In today’s world, attitudinal and social economic components such as snooze, anxiety, nationality and hormonal changes all play a role in the rise of obesity. There was a demand to establish a system that took into account the tangible, domestic, intellectual, mental and sentimental components that help to overweight, as well as top reasons to help people resolve these issues [1]. This review highlights the large body of current papers on advanced machine learning models for obesity prognostication, offering a unified view of the current schemes’ constraints. The available literature on overweight forecasting is classified into techniques that predicts consequences, characteristics used, data source types and associated purposes for the purpose of assessing the condition [2]. Obesity is an increasingly common issue that can lead to pathological conditions, lymphoma and cardiovascular disease. Obesity is influenced by a number of factors, including aging, bodyweight, tallness and BMI [3]. Numerous individuals are unsure that by making simple changes to their

9 Probable Deviation Outlier-Based Classification of Obesity with Eating …

83

wellness, they can prevent many additional problems. India’s increasingly professional life environment leads to abnormal physiological abnormalities, especially in the younger population. The government’s lenient dietary limitation policy makes harmful, refined sugar readily available [4]. Obesity rates is a significant research topic because excess weight has a negative impact on a child’s health. Especially fat and overweight newborns, early detection of childhood obesity is critical. Obesity in childhood over the age of two years can be predicted using a machine learningbased prediction model [5]. Prevalence of obesity can be predicted using machine learning algorithms in early stages of development. To predict a patient’s BMI, we created models based on information from 2 to 8 patient interactions. Our simulations reveal that for both boys and girls, five historical exposures are adequate to forecast BMI prior to the age of four [6]. Baked products and cereals are the most important dietary categories for predicting overweight, followed by dairy and sweetened beverages. The system has been tested in terms of actual confidence interval as well as the proportion of nations for which it correctly predicted obesity rates. Consumption and overweight incidence statistics from 79 countries were used in the research [7]. Obesity is associated with a greater risk of chronic illness cause of death and disability over the world. The conventional prediction model restrict the determinants that can be studied [8]. Genetic markers are found in the acquired participant characteristics using advanced analytics approaches and then categorized as risk variables in the National Human Genome Research Institute Catalog. Numerous machine learning techniques for the forecasting of obesity use annotated genetic markers as input data [9]. Classifiers always had the opportunity to assist develop appropriate dietary guidance and obesity medications. They can detect patterns of factors that are predictive of desired outcome statistically. In comparison to traditional epidemiological methodologies, machine learning algorithms can increase risk prediction for health outcomes [10]. This study attempts to determine obesity in children beyond the age of two, utilizing only evidence gathered by a healthcare delivery system called CHICA before to child’s second birthday [11]. Obesity and overweight are considered legitimate medical problems because they increase the chance of developing a variety of ailments. This research created a supervised learning model for detecting people who are obese [12].

9.3 Methodology The overall architecture of the work is shown in Fig. 9.1. The following contributions are provided in this work. . The obesity dataset contains 17 features and 2111 patient details expressed in Fig. 9.2 and it is preprocessed with encoding and missing values. . To analyze the performance metrics, the original dataset is used with all classifiers, both with and without feature scaling.

84

M. Shyamala Devi et al. Obesity Dataset

Handling-missing values

Categorical Encoding

Exploratory Data Analysis of Classes

Outlier Analysis of dataset

Outlier removed for age using STD

STD

IQR

Age

Weight

Outlier removed for Weight using STD

Outlier removed for age, weight using STD

No Feature Scaling

Outlier removed for age using IQR

Outlier removed for weight using IQR

Outlier removed for age, Weight using IQR

Feature Scaling

Fitting to logistic regression, KNN, Kernel SVM, Naive bayes, Decision tree and random forest classifier

Classification Output

Comparison of Precision, Accuracy, Recall, FScore and Run Time

Fig. 9.1 Architecture system workflow

9 Probable Deviation Outlier-Based Classification of Obesity with Eating …

85

X1 – Gender X2 - Age X3 - Height X4 - Weight X5 - Family history with_overweight X6 - FAVC - Frequent consumption of high caloric food X7 - FCVC - Frequency of consumption of vegetables X8 - NCP - Number of main meals X9 - CAEC - Consumption of food between meals X10 - SMOKE Habit X11 - CH2O - Consumption of water daily X12 - SCC - Calories consumption monitoring X13 - FAF - Physical activity frequency X14 - TUE - Time using technology devices X15 - CALC - Consumption of alcohol X16 - MTRANS - Transportation Used Y - Obesity Level (0 - Insufficient_Weight, 1 - Normal_Weight, 2- Obesity_Type_I, 3 - Obesity_Type_II, 4 - Obesity_Type_III, 5- Overweight_Level_I, 6 - Overweight_Level_II)

Fig. 9.2 Obesity level dataset feature input and output

. The exploratory data analysis is done to figure out how the target obesity variable is distributed. . The dataset is analyzed with the scatterplot to detect the features that are having outliers. From the implementation, it is observed that the features age and weight have outliers. . Dataset is applied with outlier detection methods like interquartile range and standard deviation to balance outliers in age and weight features. . The outlier is removed to form the following dataset as age outlier removed using STD, age outlier removed using IQR, weight outlier removed using STD, weight outlier removed using STD, age and weight outlier removed using STD, age and weight outlier removed using IQR. . The above outlier removed datasets are applied with various classifiers to predict the obesity levels before and after feature scaling to analyze the performances indices precision, recall, f score and accuracy.

9.4 Implementation Setup The outlier identification is done using IQR and standard deviation. Inter Quartile Range will be used to find out whether there are any outliers and is shown in Fig. 9.3 and Eq. (9.1). IQR = Q3 − Q1

(9.1)

86

M. Shyamala Devi et al.

Fig. 9.3 Representation of outlier with IQR

where Q3 is the third quarter of the information in the dataset and Q1 is the first quarter of the information in the dataset. Individual numbers that deviate from a data set’s broad pattern are known as outliers. In standard deviation method, the mean and standard deviation of the residuals are calculated and compared. A data point is considered an outlier if it deviates from the mean by a specific amount of standard deviations. The threshold is the stated number of standard deviations. /∑ n 2 i=1 (X i − X ) (9.2) SD = n−1 where ‘n’ is the total number of data points, X i is the value of each data information in the dataset. The obesity dataset with 2111 patient rows and 17 feature attributes from KAGGLE repository is subjected with the data preprocessing by eliminating the missing values and categorical values. The target obesity level distribution is analyzed and is shown in Fig. 9.4. The outlier is identified for the age and weight features in the dataset and is shown in the Fig. 9.5 and the original dataset before removing the outlier is applied with all classifiers and is shown in Table. 9.1. The age and weight outlier removed dataset using IQR is applied with all the classifiers and the performance is analyzed and is shown in Table. 9.2. The age outlier removed dataset using STD is applied with all the classifiers and the performance is analyzed and is shown in Table. 9.3. The weight outlier removed dataset using STD is applied with all the classifiers and the performance is analyzed and is shown in Table. 9.4. The age and weight outlier removed dataset using STD is applied with all the classifiers and the performance is analyzed and is shown in Table. 9.5.

9 Probable Deviation Outlier-Based Classification of Obesity with Eating …

87

Fig. 9.4 Obesity level distribution of obesity dataset

Fig. 9.5 Outlier distribution of age and weight of obesity dataset

9.5 Conclusion This paper explores the performance of obesity level classification by analyzing the outlier features in the dataset. The obesity dataset is inspected to explore outlier data distribution in the dataset and is found to have outlier in the age and weight feature. This paper is implemented to examine the obesity level target classification to prove how well the classifier accuracy is improves by removing the outliers from the age and weight features. The dataset is applied with the interquartile range and standard deviation method to remove the outliers from age and weight. Experimental results shows that the Decision Tree classifier shows accuracy of 94% before removing outliers on age and weight feature. The same Decision Tree classifier shows the accuracy of 96% after removing outliers on age and weight feature. The future work of this paper is analyze the performance of obesity level prediction by incorporating oversampling of the target.

FScore

Accu

0.62

0.95

0.96

GNB

Dtree

RFor

0.96

0.94

0.60

0.62

0.89

0.89

0.64

KNN

0.89

0.96

0.94

0.55

0.62

0.96

0.94

0.60

0.62

0.89

0.63

RunTime

0.83

0.02

0.00

0.12

0.04

0.53

0.96

0.95

0.58

0.90

0.80

0.90

0.96

0.94

0.57

0.89

0.81

0.90

Recall

Precision

0.63

0.63

Precision

0.64

Rcall

After feature scaling

Before feature scaling

KSVM

LReg

Classifier

Table 9.1 Classifier performance indices before and after feature scaling FScore

0.96

0.94

0.50

0.90

0.81

0.90

Accu

0.96

0.94

0.57

0.89

0.81

0.90

RunTime

0.88

0.02

0.00

0.13

0.05

0.12

88 M. Shyamala Devi et al.

FScore

Accu

0.76

0.88

0.93

GNB

Dtree

RFor

0.92

0.88

0.70

0.41

0.88

0.87

0.20

KNN

0.87

0.92

0.87

0.68

0.27

0.92

0.88

0.70

0.41

0.88

0.79

RunTime

0.18

0.00

0.00

0.02

0.00

0.03

0.93

0.88

0.79

0.93

0.89

0.85

0.92

0.88

0.73

0.92

0.90

0.86

Recall

Precision

0.79

0.77

Precision

0.77

Rcall

After feature scaling

Before feature scaling

KSVM

LReg

Classifier

Table 9.2 Classifier performance of age and weight outlier removed dataset using IQR FScore

0.92

0.87

0.71

0.92

0.89

0.85

Accu

0.92

0.88

0.73

0.92

0.90

0.86

RunTime

0.18

0.00

0.00

0.01

0.00

0.02

9 Probable Deviation Outlier-Based Classification of Obesity with Eating … 89

FScore

Accu

0.70

0.94

0.95

GNB

Dtree

RFor

0.94

0.94

0.68

0.87

0.90

0.91

0.91

KNN

0.90

0.94

0.94

0.66

0.88

0.94

0.94

0.68

0.87

0.90

0.44

RunTime

0.49

0.01

0.01

0.07

0.02

0.09

0.95

0.94

0.69

0.91

0.86

0.91

0.94

0.94

0.67

0.90

0.86

0.91

Recall

Precision

0.44

0.41

Precision

0.41

Rcall

After feature scaling

Before feature scaling

KSVM

LReg

Classifier

Table 9.3 Classifier performance of age outlier removed dataset using STD FScore

0.94

0.94

0.65

0.90

0.85

0.91

Accu

0.94

0.94

0.67

0.90

0.86

0.91

RunTime

0.45

0.01

0.00

0.08

0.03

0.09

90 M. Shyamala Devi et al.

FScore

Accu

0.70

0.96

0.94

GNB

Dtree

RFor

0.92

0.96

0.71

0.87

0.91

0.92

0.91

KNN

0.91

0.93

0.96

0.70

0.88

0.92

0.96

0.71

0.87

0.91

0.42

RunTime

0.48

0.01

0.00

0.07

0.02

0.10

0.94

0.96

0.61

0.92

0.81

0.89

0.92

0.96

0.62

0.90

0.82

0.89

Recall

Precision

0.42

0.37

Precision

0.39

Rcall

After feature scaling

Before feature scaling

KSVM

LReg

Classifier

Table 9.4 Classifier performance of weight outlier removed dataset using STD FScore

0.92

0.96

0.57

0.90

0.81

0.89

Accu

0.92

0.96

0.62

0.90

0.82

0.89

RunTime

0.45

0.01

0.00

0.08

0.03

0.09

9 Probable Deviation Outlier-Based Classification of Obesity with Eating … 91

FScore

Accu

0.65

0.94

0.96

GNB

Dtree

RFor

0.95

0.94

0.63

0.58

0.89

0.89

0.59

KNN

0.88

0.96

0.93

0.60

0.57

0.95

0.94

0.63

0.58

0.89

0.69

RunTime

0.56

0.02

0.00

0.09

0.02

0.09

0.96

0.94

0.65

0.91

0.85

0.88

0.95

0.94

0.63

0.91

0.85

0.88

Recall

Precision

0.69

0.67

Precision

0.68

Rcall

After feature scaling

Before feature scaling

KSVM

LReg

Classifier

Table 9.5 Classifier performance of age and weight outlier removed dataset using STD FScore

0.96

0.93

0.60

0.91

0.85

0.87

Accu

0.95

0.94

0.63

0.91

0.85

0.88

RunTime

0.51

0.01

0.00

0.08

0.03

0.09

92 M. Shyamala Devi et al.

9 Probable Deviation Outlier-Based Classification of Obesity with Eating …

93

References 1. Pereira, N.C., D’souza, J., Rana, P., Solaskar, S.: Obesity related disease prediction from healthcare communities using machine learning. In: The proceedings of the International Conference on Computing, Communication and Networking Technologies, pp. 1–7 (2019). https://doi.org/ 10.1109/ICCCNT45670.2019.8944798 2. Siddiqui, H.: A survey on machine and deep learning models for childhood and adolescent obesity. IEEE Access 9, 157337–157360 (2021) 3. Jindal, K., Baliyan, N., Rana, P.S.: Obesity prediction using ensemble machine learning approaches. In: The Proceedings of 5th ICACNI, vol. 2 (2018) 4. Lakshmanaprabu, S.K.: Effective features to classify big data using social Internet of Things. IEEE Access 6, 24196–24204 (2018) 5. Chatterjee, K., Jha, U., Kumari, P., Chatterjee, D.: Early prediction of childhood obesity using machine learning techniques. In: Advances in Communication and Computational Technology, pp. 1431–1440 (2021) 6. Cheng, E.R., Steinhard, R., Ben Miled, Z.: Predicting childhood obesity using machine learning: practical considerations. BioMedInformatics 2(1), 184–203 (2022) 7. Dunstan, J., Aguirre, M., Bastías, M., Nau, C., Glass, T.A., Felipe, D.: Predicting nationwide obesity from food sales using machine learning. Health Inf. J. (2019) 8. Thamrin, S.A., Arsyad, D.S., Kuswanto, H., Lawi, A., Nasir, S.: Predicting obesity in adults using machine learning techniques: an analysis of Indonesian basic health research. Front. Nutr. 8, 669155 (2021) 9. Montanez, C.A.C., Fergus, P., Hussain, A., Al-Jumeily, D., Abdulaimma, B., Hind, J.: Machine learning approaches for the prediction of obesity using publicly available genetic profiles. In: The Proceedings of the International Joint Conference on Neural Networks, pp. 2743–2750 (2017) 10. Selya, A.S., Anshutz, D.: Machine learning for the classification of obesity from dietary and physical activity patterns. Adv. Data Anal. Health 77–97 (2018) 11. Dugan, T.M., Mukhopadhyay, S., Carroll, A., Downs, S.: Machine learning techniques for prediction of early childhood obesity. Appl. Clin. Inform. 6(3), 506–520 (2015) 12. Rodrígueza, E., Rodrígueza, E., Nascimento, L., da Silva, A., Marinsa, F.: Machine learning techniques to predict overweight or obesity. In: The Proceedings of the International Conference on Informatics and Data-Driven Medicine (2021)

Chapter 10

Design and Implementation of Smart Home Automation System Using the Proteus Design Tool L. Niranjan, Husna Tabassum, B. Sreekantha, T. Pushpa, and Mantri Gayatri Abstract The idea of delivering home-based computerization is not a new trend in smart home technology but has been thrown into the forefront recently. Lighting, heating, air conditioning, and security are all controlled and automated. Wi-Fi is frequently utilized for controlling and remote monitoring of utmost devices. We use the Internet to monitor and operate the system via a server. It serves as a gateway to a centralized hub, from which a system may be controlled via a graphical user interface. During the operation of the system, the condition of the equipment is monitored, and the same data is displayed and included on the LCD screen for analysis purposes if the condition of the devices changes. The server also receives the same information. The system is incorporated with three sensors: a FIRE sensor, GAS sensor, and PIR sensor to detect intrusion, fire, and LPG gas detection on the premises. In addition, whenever any of the sensors is triggered, the data is communicated to the possessor through a GSM dial-up Internet. Individually, sensor takes its significance, and the appropriate action is conducted based on the state.

10.1 Introduction Labor-saving devices with gas or electric-powered domestic appliances were integrated into early home automation systems, and subsequently, in 1957, home automation system technologies with an electronic system were established. Control and L. Niranjan (B) Department of ECE, CMR Institute of Technology, Bangalore, India e-mail: [email protected] H. Tabassum · T. Pushpa Department of CSE, HKBK College of Engineering, Bangalore, India B. Sreekantha Department of ISE, HKBK College of Engineering, Bangalore, India M. Gayatri Department of CSE, Malla Reddy College of Engineering and Technology, Hyderabad, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_10

95

96

L. Niranjan et al.

automate every appliance in-house, as well as those that are far away. Imagine being able to manage the temperature of the room from your place, changing the brightness and heating of the light, or chilling the room in less than 5 min before entering it. The majority of automated systems provide security, ensuring the safety of the place. From a security camera to a water heating system that may immediately intimate you about the status of the system [1]. Your home property will be under constant observation so you can react quickly. A smartphone is used to control the device which is connected to the server via the Internet. Each piece of information is updated in a fraction of a second so that microcontroller can react if any event occurs. The same can be monitored via a system far away from the premises. Here, the system can control eight devices and monitor three sensors. Each sensor information is updated for its threshold value, if it reaches the value, action will be taken to solve that problem via the microcontroller. Let us say if a gas sensor is used, it can monitor the gas leakage in the air [2]. In the same way, a PIR sensor is used to identify if any human radiation is present on the premises in the absence of the owner and finally but not least FIRE device is accustomed to identify the fire and temperature within the residence If so, the microcontroller will act to shut off the fire by switching on the water sprinkler system. Twenty-first century homes are self-controlled and automated, providing more comforts. The existing home-based computerization system is created on wellestablished communication across a wire. The issue does not occur until and unless the system is designed and assembled during the building’s construction activities. The cost goes high for already existing buildings [3, 4]. The wireless system is used in our day-to-day life and provided great help for automation systems.

10.2 Literature Survey An innovative smart house design that incorporates WSNs and biometric technology. The technology uses biometrics for home entry authentication, which improves home security and makes it easier to walk through the door [5]. The research paper concludes with a vision for the smart home’s future when biometric technology is used more broadly and comprehensively. Line communication allows a person to operate several gadgets in a home remotely [6]. It has a smartphone app, a portable wireless remote, and a computer-based program to provide the consumer control over the devices. It gives a basic understanding of how to use a smartphone to manage and secure various household appliances [7]. The microcontroller is often used in conjunction with a mobile phone running Android as the operating system. The user may communicate with the Android smartphone by sending a command to the microcontroller, which will then operate other appliances or gather data from the sensors [8]. The technology allows you to access the device’s status and control it from a distance. This is accomplished with the assistance of the Internet. This program allows customers to remotely access their gadgets from anywhere [9]. A microcontroller is indeed a low-cost, versatile, simple-to-use fully accessible programmed microcomputer that may be utilized in a range of electrical applications. This circuit

10 Design and Implementation of Smart Home Automation System Using …

97

can be connected to any other microcontroller, microcontroller shields, and Raspberry Pi circuits to control relays, LEDs, servo motors, and actuators. The gas sensor circuit contains a USB connection, which indicates that it is possible to link a computer using USB and then configured it by using microcontroller IDE software [10]. The component has a 32 KB of flash memory which is being used to keep track of how many instructions have been given, with a 1 KB EEPROM and a 2 KB SRAM. The operating voltage of the unit is 5 V, suggesting that maybe the MCU on the PCB and accompanying microchip technology operates at that voltage, but the voltage level spans from 6 to 20 V, with a suggested range of 7 to 12 V [11]. We should know the amount of gas in the cylinder in advance of scheduling a refilling from a distributor, and the level of gas existing in the cylinder must be continuously checked for this reason [12, 13]. A load cell with the requisite weighing capacity for a residential cylinder is used, and a weighing device module is used in conjunction with the load cell for calibration purposes. The system enforces the L6D weight device module. The load cell signal powers a relay circuit, which generates two logic pulses (for = 7 kg and = 0.5 kg) that are coupled to the MCU port pins to determine the gas level [14].

10.3 Proposed Model The major intention of this research is to create also execute home computerization utilizing SCADA and the Internet of Things that can overcome issues such as high price, poor manageability, security, and inflexibility while also automating household appliances via the server [15]. The Atmel microcontroller is at the core of the abovementioned system. It is in control of all the devices that are linked to the system. It also senses any changes in the surroundings through the use of sensors that detect events. The block diagram includes an AT89C51 microprocessor, an MQ-4 gas sensor, a load cell-L6D weight switch sensor, a SIMCOM 900 GSM module, and a display (s). The system is built on the 89c52 microcontroller. The output of the gas sensor MQ-4 and the load cell L6D is the microcontroller inputs [16]. The microcontroller output is sent to the SIMCOM 900 and the LCD 162 display. The idea is implemented using a microcontroller module, however, the chips we are utilizing are made by Atmel company. Only the software or assembly code produced for the microcontroller will have a direct influence [17]. As seen from the block diagram in Fig. 10.1, there are three sensors used to detect the event. Three events have to be considered here, one is intruder detection, next is fire detection, and finally gas leakage detection. The PIR sensor is used to detect if any unauthorized user is present at home when no one is in that place. As soon as it detects, it triggers the microcontroller informing us that somebody has entered the premises [18]. In the first implementation, a push button is used to generate the logic level and the respective output is noted. A 16 × 2 LCD is used and a small message is displayed in the 1st implementation. Further, the required message is changed and this is invoked when a sensor detects the GAS molecules in the surrounding air. Furthermore, the PIR sensor is used to detect an

98

L. Niranjan et al.

intruder is present in the absence of the owner, later a fire sensor is also used to sense any fire due to a short circuit or due to accident. Here, we are achieving three types of information which are from the three sensors [7, 19]. In the first case, the MQ-4 sensor is used to detect the LPG gas density in the surrounded area. As soon as the density of air and gas ratio changes, the MQ-4 detects and triggers the microcontroller and in turn, it switches the relay driver which in turn activates the exhaust fan to remove the gas from the surrounding. The same information is fed as an SMS and sent to the user. In the second case, the PIR sensor is used to detect intruders and sends the signals to the microcontroller which takes the decision and the same will be displayed on the LCD along with an SMS alert to the user. In the third case, the fire sensor can detect any fires that may occur in the building and activates a water valve to spray water where the fire is present, this is done by the fire sensors deployed as well as the water spraying system which is present in the premises and then it sends the information to the user via SMS.

Fig. 10.1 Proposed model

10 Design and Implementation of Smart Home Automation System Using …

99

10.4 Results and Discussions There are three conditions to be considered in the implementation of this system. In the first case, the gas leakage is detected by the gas sensor, then the gas valve is activated to stop the gas leakage. The flow diagram shown in Fig. 10.2 shows how the simple sensor detects leakage of gas from the sensor, and the same is sent to the microcontroller for further action, an SMS is sent and the exhaust fan is switched ON. As soon as the gas leak is detected by the gas sensor, the port 2-bit P1.0 becomes low indicating the presence of gas as seen in Fig. 10.3. In the second case, the PIR sensor detects the presence of an intruder on the premises, where the sensor senses for the sensor output changes are compared with the predefined value which we call the threshold value. The flow diagram of the event is shown in Fig. 10.4. The trigger pulse from port P2.1 is shown in Fig. 10.5. All the actions are updated via GSM to the user. In the third case, the fire sensor detects whether there is a fire in the building as shown in the flow diagram in Fig. 10.6 and the port activation in Fig. 10.7. As soon as the controller has triggered, the water spraying system is activated. This is achieved by having the number of different nodes of the sensor and the spraying units. Fig. 10.2 Flow diagram for the gas leakage detection

100

L. Niranjan et al.

Fig. 10.3 Gas detection situation

Fig. 10.4 Flow diagram for PIR sensor

The design comprises a microcontroller where the user can relate with the system and can refer switch signals to the controller this, in turn, is in charge of controlling various embedded systems/sensors. As soon as the microcontroller is powered ON, the message “WELCOME TO HOME AUTOMATION” is displayed on the LED as shown in Fig. 10.8. Figure 10.9 shows the gas leakage in the premises and the same information is sent via the GSM module. As shown in Fig. 10.10, the LCD

10 Design and Implementation of Smart Home Automation System Using …

101

Fig. 10.5 PIR sensor detection simulation

Fig. 10.6 Flow diagram for fire sensor

displays the condition of the intruder which is sensed by the sensor. This project uses a passive infrared sensor [PIR sensor] to identify the intruder on the premises, which is linked to port pin 2.5 of the microprocessor. The main goal is to create an alarm-based safety system that uses power as an input and uses a motion sensor to detect motion and display a message on an LCD saying “INTRUDER DETECTED

102

L. Niranjan et al.

Fig. 10.7 FIRE sensor simulation

CHECK IMMEDIATELY” and this message will be updated to the person by SMS through GSM system. As shown in Fig. 10.11, the FIRE detection is detected by using a fire detection sensor. Table 10.1 gives the simulation time and real-time sensor and other modules reacting time. The entire circuit is designed using the proteus design tool which is shown in Fig. 10.12, and the real-time setup of the entire system is shown in Fig. 10.13, which includes the controller, relay module, GSM module, and LCD. Fig. 10.8 Initial condition and normal mode

10 Design and Implementation of Smart Home Automation System Using …

103

Fig. 10.9 GAS leakage mode

Fig. 10.10 PIR detection mode

10.5 Conclusion This research paper describes our proposed system for gas monitoring, leakage, automatic booking, and manual booking. The project’s main goal is to decrease the deaths that occurred due to the leakage of gas from the cylinder as well as negligence by the users. Here, the system controls the flow of gas from the cylinder to the stove or any other devices in need of the LPG gas. As soon as the system comes to know if any gas leakage, it will automatically shut off the gas cylinder valve by using a solenoid valve, if this fails in case, the controller will immediately turn on the exhaust fan, to eliminate the gas from the surrounding place. The second advantage of this system is it can book a new cylinder as soon as the cylinder weight is less than 20% of its total weight. The third advantage is it gives a convenient way for the operator

104

L. Niranjan et al.

Fig. 10.11 FIRE detection mode

Table 10.1 Simulation versus time S. No.

Task

1

Sensors

2

Type

Simulation results

Real-time Results

Difference

Gas MQ2

0.74S

1.21S

0.47S

PIRE

0.72S

1.44S

0.72S

FIRE

0.82S

1.32S

0.5S

3

Display

LCD 16 × 2

1.3S

1.75S

0.45S

4

Buzzer

Piezoelectric

1.2S

1.73S

0.53S

5

Relay

12 V relay

3.2S

4.56S

1.36S

3

Fig. 10.12 PCB layout of the entire module

10 Design and Implementation of Smart Home Automation System Using …

105

Fig. 10.13 Proposed system hardware module snapshot

to manually book the cylinder by just clicking on a button. Further, the entire circuit is designed and tested using the proteus design tool. Along with this, the code is written using Keil-c software. The final module PCB is designed using the proteus design tool as seen in this research paper. Furthermore, the entire data is reorganized in the cloud server where the data is gathered whenever it is necessary.

References 1. Gurek, A., Gur, C., Gurakin, C., Akdeniz, M., Metin, S.K., Korkmaz, I.: An android based home automation system. IEEE (2013) 2. Hasan, A., Khan, A.A.M., Uddin, N., Mitul, A.F.: Design and implementation of touchscreen and remote control based home automation system. In: Proceedings of 2013 2nd International Conference on Advances in Electrical Engineering (ICAEE 2013), Dhaka, Bangladesh (2013) 3. AlShu’eili, H., Gupta, G.S., Mukhopadhyay, S.: Voice recognition based wireless home automation system. In: 4th International Conference on Mechatronics (ICOM), Kuala Lumpur, Malaysia (2011) 4. Rosendahl, A., Hampe, J.F., Botterweck, G.: Mobile home automation—merging mobile value added services and home automation technologies. In: Sixth International Conference on the Management of Mobile Business (ICMB) (2007) 5. Chintalapati, J.B., Rao, S.T.Y.S.: Remote computer access through Android mobiles. 9(5), 3 (2012) 6. Singh, N., Bharti, S.S., Singh, R., Singh, D.K.: Remotely controlled home automation system. In: IEEE International Conference on Advances in Engineering and Technology Research (ICAETR) (2014) 7. Shwetha, N., Niranjan, L., Chidanandan, V., Sangeetha, N.: Smart driving assistance using Arduino and Proteus design tool. In: Expert Clouds and Applications, pp. 647–663 (2021).https://doi.org/10.1007/978-981-16-2126-0_51 8. Liu, Y.: Study on smart home system based on Internet of Things technology. In: Du, W. (ed.) Informatics and Management Science IV, vol. 207, pp. 73–81. Springer, London (2013) 9. Priyatham, M.M.: Lifetime ratio improvement technique using special fixed sensing points in wireless sensor network. Int. J. Pervasive Comput. Commun. 17(5), 483–508 (2021).https:// doi.org/10.1108/IJPCC-10-2020-0165 10. Al-Qutayri, M.A., Jeedella, J.S.: Integrated wireless technologies for smart homes applications. In: Al-Qutayri, M.A. (ed.) Smart Home Systems. InTech (2010)

106

L. Niranjan et al.

11. Chung, C.-C., Huang, C.Y., Wang, S.-C., Lin, C.-M.: Bluetooth-based Android interactive applications for smart living. In: Second International Conference on Innovations in Bio-inspired Computing and Applications (IBICA), pp. 309–312 (2011) 12. Aware, A., Vaidya, S., Ashture, P., Gaiwal, V.: Home automation using Android App and cloud network. Int. J. Eng. Res. Gen. Sci. 3(3) (2015) 13. Niranjan, L., Priyatham, M.M.: An energy-efficient and lifetime ratio improvement method based on energy balancing. Int. J. Eng. Adv. Technol. 9(1S6), 52–61 (2019). https://doi.org/ 10.35940/ijeat.a1012.1291s619 14. Hasan, M., Anik, M.H., Islam, S.: Microcontroller based smart home system with enhanced appliance switching capacity. In: HCT Information Technology Trends (ITT) 2018 Fifth, pp. 364–367 (2018) 15. Shwetha, N., Niranjan, L., Chidanandan, V., Sangeetha, N.: Advance system for driving assistance using Arduino and proteus design tool. In: Third International Conference on Intelligent Communication Technologies and Virtual Mobile Networks (ICICV) (2021). https://doi.org/ 10.1109/icicv50876.2021.9388620 16. Ibrahim, S., Shukla, V.K., Bathla, R.: Security enhancement in smart home management through multimodal biometric and passcode. In: International Conference on Intelligent Engineering and Management (ICIEM), pp. 420–424 (2020) 17. Gladence, L.M., Sangeetha, K.K., Soundharya, S., Revathy, S., Selvan, M.P.: Smart home monitoring system and prediction of power consumption. In: 6th International Conference on Computing Methodologies and Communication (ICCMC), pp. 1–7 (2022) 18. Narayanan, V.S., Gayathri, S.: Design of wireless home automation and security system using PIC microcontroller. Int. J. Comput. Appl. Eng. Sci. 3, 135–140 (2013) 19. Ahmad, A.M., Jan, N., Iqbal, A., Lee, C., Korea, A.: Implementation of ZigBee-GSM based home security monitoring and remote control system. In: IEEE Symposium on Circuits and Systems (MWSCAS), Seoul, pp. 1–4 (2011)

Chapter 11

A Novel Approach to Prognosticate CKD Using a Supervised and Unsupervised Learning Algorithms S. Ashwathi, A. Swamy Goud, L. Niranjan, B. Sreekantha, and J. Suneetha

Abstract Chronic kidney disease (CKD) occurs when your kidneys get spoiled and therefore unable to purify the blood as efficiently as they should. The ailment is termed “chronic” because the impact on your kidneys occurs progressively over time. As a result of the injury, waste can accumulate in your body. CKD can lead to a variety of other health problems. To diagnose CKD in its initial stages, a variety of approaches and technologies have been proposed. Machine learning (ML) technologies are especially important in the early diagnosis of a range of diseases. This study used five ML algorithms: KNN, CHIRP, J-48 decision tree, random forest, and deep belief network. The function of the dataset determines the efficiency of classification technologies. An algorithm model has been designed to increase the categorization system’s efficiency by lowering the variable dimension. Furthermore, the accuracy results of the experiments indicated 99.75% for CHIRP, 97.3% for KNN, 100% for J-48 decision tree, 99.63% for random forest, and 98.5% for DBN. Overall, the J-48 decision tree outperforms other decision trees when it comes to reducing inaccuracy rates and enhancing precision.

S. Ashwathi (B) · J. Suneetha Department of CSE, CMR Institute of Technology, Bangalore, India e-mail: [email protected] A. Swamy Goud Department of IT, BVR Institute of Technology, Narsapur, Hyderabad, India L. Niranjan Department of ECE, CMR Institute of Technology, Bangalore, India B. Sreekantha Department of ISE, HKBK College of Engineering, Bangalore, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_11

107

108

S. Ashwathi et al.

11.1 Introduction Chronic kidney disease (CKD) is growing progressively popular and well known as a global issue. This is a long-term issue that has been associated with greater mortality and doom, as well as a higher chance of various other disorders including higher human service and heart disease expenses [1]. CKD also referred to as a chronic renal failure affects so many individuals that it often goes unreported and unidentified until the disease has advanced to a severe degree. CKD testing is routinely performed on individuals with diabetes, hypertension, smoking, high blood pressure, alcohol intake, ethnicity, heart disease, age, obesity, and those people that have CKD in their family. Machine learning methods have recently been found to help assess illness in the prior phase. The most common approach for identifying renal disease is a machine-predicted analysis. As a consequence, there are a few adverse symptoms that CKD may not be noticeable until significant renal function is hampered [2, 3]. The goal of CKD therapy is to slow the course of renal disease by addressing the root cause of the disease early on the epidemiology which has a role in the development of CKD, as well as many other clinical features. In general, nephrologists utilize blood tests and urine testing to see whether you have chronic kidney disease. Genetics, diabetes, obesity, and age are all factors that might contribute to CKD [4, 5]. The current research contributes to the creation of an ML-based method for identifying CKD. The following is a summary of the paper’s contributions: (a) KNN, J-48 decision tree, CHIRP, KNN, DBN, and random forest were five machine learning algorithms employed to detect CKD with promising accuracy. (b) Expert physicians can assist popularize highly effective machine learning approaches for the detection of long-lasting renal illness. Figure 11.1 depicts the normal pattern of CKD diagnosis. Fig. 11.1 Causes of CKD

11 A Novel Approach to Prognosticate CKD Using a Supervised …

109

11.2 Related Works The blood test determines how successfully creatinine, a byproduct of routine muscle breakdown, is eliminated from the blood through the kidneys [6]. The urine test, on the other hand, will show that protein is present in the urine. Protein (albumin) is indeed a bloodstream particle that is generally not passed through the kidney filter and into the urine [7]. The appearance of protein in the urine is a sign that something is wrong that the kidney channels have been destroyed and may suggest chronic kidney illness. Using machine learning techniques such as KNN, CHIRP, random forest, J-48 decision tree, and an AHDCNN, variety of experiments were run on the CKD dataset [8, 9]. In our study [10], the mean methodology was obtained by calculating missing numerical values during preprocessing, whereas the prototype system was used to calculate misplaced minimal values [11]. The pertinent aspects of the critical features for chronic kidney disease analysis were chosen using the recursive feature elimination technique [12]. These traits were used to train predictors that could diagnose illnesses. The random forest, KNN, CHIRP, and J-48 decision tree algorithms were used to cure chronic kidney disease in this work [13]. All of the algorithms performed though when it got to identifying whether a patient showed CKD or a good kidney. To handle classification problems, several researchers employ K-nearest neighbor (KNN). On a dataset derived from Pima Indians data, employed KNN for CKD. KNN predicts CKD with an accuracy of 76.96% and a minimum error rate, according to their findings [14]. Based on unlabeled data, unsupervised algorithms can make inferences or predictions [15]. To overcome the difficulties of nonlinear recognizability, computational unpredictability, and dimensionality, Composite Hypercube on Iterated Random Projection (CHIRP) is a repetitive element consisting of 3 stages: planning, coating, and grouping.

11.3 Proposed System The datasets of the preprocessing are applied to the reference model shown in Fig. 11.1, where 400 participants in the CKD material source came from the ML Archives at the California State University, Irvine, were used. In addition to specific attributes which include “CKD” and “non-CKD,” the dataset contains 24 aspects, 11 numeric aspects, and 13 categorical aspects. Blood pressure, RBC count, sugar, blood sugar random, sodium, packed cell volume, hemoglobin, potassium, WBC count, coronary vein disease, pus cell, RBC, hypertension, diabetes mellitus, albumin, anemia, pus cell clusters, serum creatinine, age, appetite, blood urea, pedal edema, bacteria, and specific gravity are just a few of the symptoms. To determine the kidney stages, we have considered major 10 datasets, and this is illustrated in Table 11.1. A strategy for transforming raw data into a clear dataset is data preparation. This stage is the starting point for any machine learning classification technique. This approach handles missing values, rescales the dataset, converts it to binary data, and

110

S. Ashwathi et al.

Table 11.1 10 dataset attributes to determine the CKD stage

S. No

Description

Attribute

Type

1

Blood pressure

bp

Discrete integer values

2

Albumin

al

Nominal values

3

Blood urea

bu

Discrete integer values

4

Serum creatinine

sc

Numeric values

5

Sodium

sod

Discrete integer values

6

Potassium

pot

Numeric values

7

Hemoglobin

hemo

Numeric values

8

Diabetes mellitus

dm

Nominal values

9

Appetite

appet

Nominal values

10

Anemia

ane

Nominal values

Features

standardizes it. The mean of the standardized method is 0 and the standard deviation is 1. In the feature selection as shown in Fig. 11.2 uses vector features which are extracted to reduce worthless and irrelevant features for prediction, preventing the building of a robust diagnostic model. The three fundamental types of feature selection methods are filter, wrap, and implanted. By choosing the best data in the dataset, a suitable feature selection increases the classifier’s performance and reduces computation time. When it comes to medical data, for example, the issue might be tough to describe why certain attributes are not included in the dataset. Each feature selection approach has its own set of limitations, making it problematic to comprehend why firm traits are picked without digging into the precise

Red Blood Cell Blood Urea Anemia_Yes Serium Creatine Pedal Edema_Yes Diabetes Specific Gravity Albumin 0 2 4 6 Feature Importance

(a) Fig. 11.2 a Reference model, b feature importance of parameters

(b)

11 A Novel Approach to Prognosticate CKD Using a Supervised …

111

invention. The dataset for all of the classifier models is the same. Each classifier, DBN, J-48 decision tree, CHIRP, KNN, and random forest, has a novel concept that was conditioned to distinguish the information into groupings of CKD and non-CKD cases, as revealed in Fig. 11.1.

11.3.1 Classifier Validation Method There were two sections to the dataset: a training dataset and a testing dataset. The partitioning and prediction of the results were done with the IBM SPSS modeler. Half of the data is in the training data group, while the remaining is in the testing data group. To transform the category of attributes, an IBM SPSS type tool was used. Each classification method received a performance evaluation matrix.

11.3.2 Measures for Performance Evaluation The performance of the classifier was checked using a variety of assessment matrices. The confusion matrix was utilized for this purpose. Due to two classes in the dataset, it is a 22 matrix. The confusion matrix shows two categories of accurate classifier predictions and two categories of inaccurate classifier predictions.

11.3.3 Description of Confusion Matrix True Positive (TP) denotes a positive output, indicating that the projected result has been accurately categorized. True Negative (TN) denotes a negative output that accurately classifies the predicted result. False Positive (FP) denotes that the outcome is positive output but the predicted outcome is classified wrongly. False Negative (FN) denotes that the result is negative, causing the projected result to be classified wrongly.

11.3.4 Classification Accuracy The right ratio of estimated findings is shown by classification accuracy from Eq. 11.1. The confusion matrix is used to calculate it. The equation is used to determine classification accuracy. Accuracy =

TN + TP FP + FN + TN + TP

(11.1)

112

S. Ashwathi et al.

11.3.5 Classification Error The proportion of inaccurately estimated outcomes is shown by the classification error from Eq. 11.2. The confusion matrix is used to calculate it. The equation is employed to find the classification error. Error Rate = FP + FN

(11.2)

11.3.6 Precision Precision is a critical criterion for evaluating model performance. It is the percentage of linked instances in the overall number of recovered cases. It is a forecasted positive rate. In the equation, precision is evaluated using Eq. 11.3. Precision =

TP FP + TP

(11.3)

11.3.7 Recall A recall is a vital criterion for evaluating model presentation. It is the proportion of associated illustrations in the overall number of cases recovered. In the equation, the recall is determined using Eq. 11.4. Recall =

TP FN + TP

(11.4)

11.3.8 F1 Score The F1 score is divided into two components: recall and precision. The F1 score aims to combine accuracy and recall measurements into a specific number. We compute the mean of accuracy and recollect in the F1 score. The harmonic mean is an obvious option because they are both rates from Eq. 11.5. F1 Score = 2 ∗

Precision ∗ Recall Precision + Recall

(11.5)

11 A Novel Approach to Prognosticate CKD Using a Supervised …

113

Fig. 11.3 a Serum creatinine accuracy analysis, b albumin accuracy analysis

11.4 Results and Discussions The accuracy analysis of serum creatinine levels in the blood for a dataset of 50 is shown in Fig. 11.3a. The albumin accuracy analysis levels in the blood for a dataset of 50 are shown in Fig. 11.3b. The accuracy analysis of potassium levels in the blood for a dataset of 50 is shown in Fig. 11.4a. The urea levels in the blood for a dataset of 50 are shown in Fig. 11.4b. The sodium levels in the blood for a dataset of 50 are shown in Fig. 11.5a. The diabetes levels in the blood for a dataset of 50 are shown in Fig. 11.5b. The dataset versus accuracy analysis of serum creatinine in the blood for a dataset of 50 is shown in Fig. 11.6. In comparison with RF, DBN, KNN, and CHIRP, the J-48 DT has a superior percentage analysis. The J-48 DT prediction algorithm has the highest accuracy during the analysis procedure.

11.5 Conclusion This study offered insight on how to detect CKD symptoms in the early phases of the ailment to battle their condition as well receive therapy. A total of 400 patients were used to create the dataset, which comprised of 24 characteristics. The data was split into two sections. 80% of the time is spent on training, while testing and validation take up 20%. As observed from the previous approaches, the CKD was determined by using 25 features and which was a tedious process and time-consuming. In the proposed approach, we have considered features, namely serum creatinine, albumin, potassium, urea, sodium, and diabetes. Using these features, we can determine the stage of CKD earlier. The outliers were removed from the dataset, and missing data was replaced. As a result, all algorithms produced encouraging outcomes. All

114

S. Ashwathi et al.

Fig. 11.4 a Potassium accuracy analysis, b urea accuracy analysis

Fig. 11.5 a Sodium accuracy analysis, b diabetes accuracy analysis

other algorithms were outperformed by the J-48 decision tree attaining F1 score correctness, precision, and recall for all measures, a score of 100% is required. The system was investigated. The results of a multiclass statistical study, as well as the CHIRP, KNN, DBN, and random forest empirical results, 99.01, 98.7, 98.5, and 99.23% were deemed to be significant in terms of the accuracy metric. The proposed system compared to the previous approach is better in determining the CKD stage. The J-48 is fast and far better compared to CHIRP, KNN, DBN, and RF in determining the CKD stage.

11 A Novel Approach to Prognosticate CKD Using a Supervised …

115

Fig. 11.6 Dataset versus accuracy analysis of serum creatinine

Acknowledgements This research was supported by SiliconShelf Electronic Systems, Bangalore, for the support during the evaluation process at CMRIT, Bangalore. We thank our colleague Prof. Niranjan L, ECE, CMRIT who provided insight and expertise that significantly aided the research. We thank Mr. Vinay for his support with methodology from Siliconshelf Electronics Systems.

References 1. Ani, R., Sasi, G., Sankar, U.R., Deepa, O.S.: Decision support system for diagnosis and prediction of chronic renal failure using random subspace classification. In: International Conference on Advances in Computing, Communications and Informatics (ICACCI) (2016). https://doi. org/10.1109/icacci.2016.7732224 2. Gupta, B., Rawat, A., Jain, A., Arora, A., Dhami, N.: Analysis of various decision tree algorithms for classification in data mining. Int. J. Comput. Appl. 163(8), 15–19. https://doi.org/ 10.5120/ijca2017913660 3. Tangri, N., Kitsios, G.D., Inker, L.A., Griffith, J., Naimark, D.M., Walker, S., Levey, A.S.: Risk prediction models for patients with chronic kidney disease: a systematic review. Ann. Intern. Med. 158(8), 596–603 (2013) 4. Owens, C.D., Ho, K.J., Kim, S., Schanzer, A., Lin, J., Matros, E., Belkin, M., Conte, M.S.: Refinement of survival prediction in patients undergoing lower extremity bypass surgery: stratification by chronic kidney disease classification. J. Vascular Surg. 45(5), 944–952 (2007) 5. Niranjan, L., Priyatham, M.M.: Lifetime ratio improvement technique using special fixed sensing points in wireless sensor network. Int. J. Pervasive Comput. Commun. 17(5), 483–508 (2021).https://doi.org/10.1108/IJPCC-10-2020-0165

116

S. Ashwathi et al.

6. Kocyigit, E.E., Unal, A., Sipahioglu, M.H., Tokgoz, B., Oymak, O., Utas, C.: Role of neutrophil/lymphocyte ratio in prediction of disease progression in patients with stage-4 chronic kidney disease. J. Nephrol. 26(2), 358–365 (2013) 7. Kolachalama, V.B., Singh, P., Lin, C.Q., Mun, D., Belghasem, M.E., Henderson, J.M., Francis, J.M., Salant, D.J., Chitalia, V.C.: Association of pathological fibrosis with renal survival using deep neural networks. Kidney Int. Rep. 3(2), 464–475 (2018) 8. Miotto, R., Li, L., Dudley, J.T.: Deep learning to predict patient future diseases from the electronic health records. In: Proceedings of the European Conference on Information Retrieval. Springer, Cham, Switzerland, pp. 768–774 (2016) 9. Shwetha, N., Niranjan, L., Gangadhar, N., Jahagirdar, S., Suhas, A.R., Sangeetha, N.: Efficient usage of water for smart irrigation system using Arduino and proteus design tool. In: 2nd International Conference on Smart Electronics and Communication (ICOSEC) (2021). https:// doi.org/10.1109/icosec51865.2021.9591709 10. Fliser, D., Kollerits, B., Neyer, U., Ankerst, D.P., Lhotta, K., Lingenhel, A., Ritz, E., Kronenberg, F.: Fibroblast growth factor 23 (FGF23) predicts progression of chronic kidney disease: the mild to moderate kidney disease (MMKD) study. J. Am. Soc. Nephrol. 18(9), 2600–2608 (2007) 11. Niranjan, L., Priyatham, M.M.: An energy efficient and lifetime ratio improvement methods based on energy balancing. Int. J. Eng. Adv. Technol. 9(1S6), 52–61 (2019). https://doi.org/ 10.35940/ijeat.a1012.1291s619 12. Surendiran, J., Saravanan, S.V., Manivannan, K.K.: Detection of glaucoma based on color moments and SVM classifier using K mean clustering. IJPT 8(3), 16139–16148 (2016) 13. Furey, T.S., Cristianini, N., Duffy, N., Bednarski, D.W., Schummer, M., Haussler, D.: Support vector machine classification and validation of cancer tissue samples using microarray expression data. Bioinformatics 16(10), 906–914 (2000) 14. Vidhya, R.G., Saravanan, R., Rajalakshmi, K.: Mitosis detection for breast cancer grading. Int. J. Adv. Sci. Technol. 29(3), 4478–4485 (2020) 15. Kim, D.-H., Ye, S.-Y.: Classification of chronic kidney disease in sonography using the GLCM and artificial neural network. Diagnostics 11(5), 864 (2021)

Chapter 12

Deep Network Architectures for Object Detection and Tracking: A Review Chinthakindi Kiran Kumar, Gaurav Sethi, and Kirti Rawal

Abstract Object segmentation is a vital topic of research in image processing and computer vision. This helps in image analysis, video surveillance, image compression, robotic and automatic process and many others. With the inclusion of deep learning methods in computer vision applications has paved a new way for object segmentation with high precision and accuracy. There is substantial amount of works that are aimed at developing effective object segmentation approaches using deep convolutional networks. This paper provides a comprehensive review on the deep learning models which are pioneered in providing effective segmentation results. The review focuses to investigate the challenges, strengths with these deep network models and also provides information on widely used datasets in research. Compare the performances of these networks and provide directive points for conducting further research.

12.1 Introduction The problem of image segmentation is vital problem in computer vision applications [1]. This segmentation process involves partitioning of images that is a crucial component in understanding the visual systems. The analysis may include medical images, commercial application like video surveillance and object tracking. The problem of image segmentation can be framed as problem of categorizing pixels with semantic labels which are key elements in partitioning the individual objects. The process of semantic segmentation involves in labeling the pixels with the set off categories like C. K. Kumar (B) · G. Sethi · K. Rawal School of Electronics and Electrical Engineering, Lovely Professional University, Phagwara, Punjab, India e-mail: [email protected] G. Sethi e-mail: [email protected] K. Rawal e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_12

117

118

C. K. Kumar et al.

human, car, sky, chair, etc. So it becomes more demanding application rather than whole image classification that aims to predict single label instead of whole image. Semantic segmentation outspreads this scope by identifying their boundaries off an object of interest in an image. In literature survey, various image segmentation algorithms were proposed like thresholding algorithms [2], K-means clustering [3], watershed algorithms [4], active content models [5], Markov models [6], sparsity-based algorithms [7] and others. In recent times, deep learning models have attained the interest of solving computer vision problems and most of the segmentation problems got results with remarkable performance reaching high accuracy on benchmark datasets. In recent times, multiple video datasets are devoted for background subtraction and foreground extraction. These datasets address the problem of realistic scenarios and meant to evaluate the performance off different algorithms. These datasets are accessible to research communities and help in building new methods and approaches and evaluate them accordingly. In traditional object detection methods, researchers have used several features to extract the background and foreground regions; these features include histogram of histogram of gradients, frame difference approach, the scale invariant feature transform and other optical flow methods. However, it was observed that these algorithms may not provide more precise result. Recent developments in deep learning models have shown tremendous outcomes in object detection and tracking. These are performed on several large datasets paving a way for future research. This survey covers insight details of recent deep learning models that aims to segment the objects and provides a brief review of these networks that includes training datasets, loss functions and training strategies. This review also summarizes the performance of different network models of multiple datasets and presents the challenges that were faced for designing the networks.

12.2 Deep Neural Network Models An overview of deep neural networks (DNN) used in computer vision community are discussed in this section.

12.2.1 Convolutional Neural Networks (CNNs) CNN network models are most successful and widely used models in deep learning exclusively for computer vision applications. These network models are initially proposed by Fukushima [8] which is followed by several others lie Waibel et al. [9], who have shared the concept of weights which are shared among temporal receptive fields. On other hand, LeCun et al. [10] have developed a fully CNN model used for document recognition. Figure 12.1 presents an overview of the CNN fundamental

12 Deep Network Architectures for Object Detection and Tracking: A Review

119

Fig. 12.1 Basic architecture of CNN model

architectural components. In general, CNN uses three types of layers (i) convolutional layers, here the weights or kernel are convolved with input samples to extract features (ii) nonlinear layers that usually applied with an activation function on the extracted feature maps (iii) pooling layers, which are meant to minimize the spatial resolution of neighborhood to enhance the feature map depth with some statistical data like mean, SD, max and others. In conventional neural network, layers locally connect with each other and receive input weights from neighborhood called as respective regions of previous layers. By integrating all these layers forms a pyramid like structure known as multi-resolution pyramids and size of it varies with number of layers. Unlike neural network, the CNN provides an advantage of sharing the weights resulting small amount of parameters than conventional networks. Some famous CNNs models include such as AlexNet [11], ResNet [12] and VGGNet [13].

12.2.2 Long Short-Term Memory (LSTM) Networks A type of recurrent neural network (RNN) is known as LSTM which is designed to overcome the issues related to long sequences dependency and gradient vanishing [14]. In general, RNNs are used to process the sequential data some of them like video sequences, speech signals, EEG signals and other time series. Figure 12.2 depicts the basic architecture of LSTM mode. At every step ‘t’, the model acquires input sample x t and hidden state information ht −1 from proceeding state and sends the output as Ot [15]. This structure contains three major layers termed as gates like input gate, forget gate and output gate, and these gates control the information flow into memory cell that store the previous state information.

12.2.3 Encoder and Decoder Model These models cram the data points from the input domain to output domain with 2 stage network. Encoder executes function z = g(x), an encoding function aims to compress input data ‘x’ into new space depiction ‘z’. This acquires semantic data of input which is used to predict the output. These are dominant for sequence

120

C. K. Kumar et al.

Fig. 12.2 Basic architecture of LSTM model

Fig. 12.3 Typical structure of encoder-decoder model

to sequence model in natural language processing (NLP) and also image-image translation where output will be improved when compare with original image like de-noising, blurring and super resolution applications [16]. Figure 12.3 depicts the encoder-decoder model structure.

12.2.4 Generative Adversarial Networks (GANs) These are new part of deep learning models which contains of two networks, one is generator and other one is discriminator. The generator network ‘G’ learns the mapping from noise ‘z’ to a target distribution ‘y’. The discriminator network ‘D’ aims to differentiate which is original and which is fake. This model can be treated as min–max distribution among ‘G’ and ‘D’ that tries to reduce the error in removing the fake samples. When loss function is maximized, then ‘G’ tries to improve the discriminator network error so that the loss function is reduced. Variety of GANs are proposed in literature like Convolutional GAN [17], Wasserstein GANs [18] and Conditional GANs [19]. Figure 12.4 illustrates the fundamental structure of GAN model.

12 Deep Network Architectures for Object Detection and Tracking: A Review

121

Fig. 12.4 Basic architecture of GAN model

12.3 Deep Learning Models for Image Segmentation There are numerous segmentation models which are based on deep learning models and these models are categorized based on their fundamental architectures. Most of these segmentation models have some similar methods such as encoderdecoder, multi-scale structures and dilated convolutions. Based on their architectural attributes, these segmentation models can be categorized as given in Table 12.1.

12.4 Deep Learning Segmentation Models Performance Analysis To assess the performance of multiple deep learning models on different datasets, in this work, several metrics were compared which gives us an insight conclusion with the selection of models. These metrics include pixel accuracy (PA), mean-intersection over union (MIOU), or Jaccard index, mean pixel accuracy (MPA), precision, recall, F1-score and dice coefficient. The above are the tabulated results that were mentioned in authorized articles and the accuracies that were obtained with different datasets and presented in those articles. This work has reviewed most of the articles pertaining to these approaches that were applied on the respective datasets and tabulated the results based on its concluding remarks. Table 12.2 gives the DL model performances on PASCALVOC datasets and it can be observed that there is a significant development in the accuracies of the models. Table 12.3 gives the performance of DL models that were attained on cityscape dataset, and it can be observed that the latest method has achieved an improvement of ~ 20% more when compared with traditional or early stage DL models.

122

C. K. Kumar et al.

Table 12.1 Deep learning models for image segmentation Category

Author

Approach

Application

Fully convolutional models

Long et al. [20]

Contains only convolutional layers. Modified existing VGG16 and GoogLeNet by removing fully connected layers

Brain segmentation [21]

Liu et al. [25]

Proposed ParseNet that adds global context to FCN by augmenting average feature

Volumetric MRI image analysis

Chen et al. [26]

Combines CNN with conditional random fields (CRF)

Can localize the segmented boundaries accurately

Urtasun et al. [27]

Combined CNN + CRF

Semantic segmentation on PASCAL-VOC dataset

Lin et al. [28]

CNN + Contextual CRF

Patch background extraction

Liu et al. [29]

Combined CNN with End to end computation can be Markov random fields done in one pass (MRF) called it as parsing network

Noh et al. [30]

Proposed DeConvNet For general purpose semantic consists of an encoder segmentation with VGG16 layers and a multi-layer de-convolutional layer

Fu et al. [31]

Stacked de-convolutional network

Chaurasia [32]

LinkNet

Xia et al. [33]

W-Net

Lin et al. [34]

Feature pyramid network merges low and high resolution features

Zhao et al. [35]

Pyramid scene parsing network where features are extracted using ResNet with dilated network

CNN with graphical models

Encode-decoder-based models

Multi-scale networks

Instance-aware-segmentation [22] Skin lesion segmentation [23] Iris segmentation [24]

Used for semantic segmentation, MRI image segmentation and object detection

Used for semantic segmentation

(continued)

12 Deep Network Architectures for Object Detection and Tracking: A Review

123

Table 12.1 (continued) Category

R-CNN-based models

Dilated convolutional models

RNN-based models

Author

Approach

He et al. [36]

Dynamic multi-scale filters network

He et al. [37]

Adaptive pyramid context network

Dollar et al. Masked R-CNN [38]

Application

Outperformed in COCO object instance segmentation challenge

Liu et al. [39]

Path aggregation network which is based on masked R-CNN and FPN

Dai et al. [40]

Multi-task for instance Outputs box detection and segmentation which semantic segmentation used consists of cascaded for object tracking structure aims for differentiating instances, estimating masks and categorizing masks

Chen et al. [41]

MaskLab, by refining object detection with semantic features based on R-CNN

PinHeiro et al. [42]

Deep mask

Xie et al. [43]

Polar mask

Lee et al. [44]

Center mask

Chen [26, 45]

DeepLab family [45]

Wang et al. [46]

Dense up-sampling convolution and hybrid dilated convolution (DUC-HDC)

Visin et al. [47]

ReSeg model which is based on ResNet whose layers are stacked which are followed by up-sampling

Byeon et al. 2D LSTM networks [48]

Texture segmentation (continued)

124

C. K. Kumar et al.

Table 12.1 (continued) Category

Attention-based models

GAN models

Table 12.2 Performance analysis of DL models on PASCAL-VOC datasets

Author

Approach

Application

Liang et al. [49]

Graph-based LSTM

Super pixel-based semantic segmentation

Huang et al. [50]

Reverse attention network (RAN)

RAN performs the direct and reverse attention learning

Li et al. [51]

Pyramid attention network

Exploits global contextual information for semantic segmentation

Fu et al. [52]

Dual attention network

Scene segmentation

Luc et al. [53]

Proposed adversarial training approach

Soult et al. [54]

Semi-weakly supervised semantic segmentation

Semantic segmentation and cell image segmentation for tissue classification

Hung et al. [55]

Semi-supervised network

Method

Backend approach

M-IOU

FCN [20]

VGG16

62.2

CRF + RNN [55]

–

72.0

BoxSup [56]

–

75.1

GCN [57]

ResNet-152

82.2

Piecewise [28]

–

78.0

DeepLab-CRF [45]

ResNet-101

79.7

RefineNet [58]

ResNet-152

84.2

Wide ResNet [59]

Wide ResNet-38

84.9

PSPNet [34]

ResNet-101

85.4

Deeplab V3 [26]

ResNet-101

85.7

PSANet [60]

ResNet-101

85.7

EncNet [61]

ResNet-101

85.9

DFN [62]

ResNet-101

86.2

SDN [31]

DenseNet-161

86.6

DIS [63]

ResNet-101

86.8

APC-NET [37]

ResNet-101

87.1

Deeplabv3 + [64]

Xception-71

87.8

Efficient Net + FPN [65]

–

90.5

12 Deep Network Architectures for Object Detection and Tracking: A Review Table 12.3 Performance analysis of DL models on cityscapes datasets

125

Method

Backend approach

M-IOU

FCN [20]

VGG16

65.3

CRF + RNN [55]

–

73.0

FoveaNet

ResNet-101

74.1

DeepLab-CRF [45]

ResNet-101

79.4

GCN [57]

ResNet-152

76.9

RefineNet [58]

ResNet-152

73.6

Wide ResNet [59]

Wide ResNet-38

84.9

PSPNet [34]

ResNet-101

85.4

Deeplab V3 [26]

ResNet-101

81.3

PSANet [60]

ResNet-101

80.1

DFN [62]

ResNet-101

79.3

Exfuse [66]

ResNet-101

86.2

SDN [31]

DenseNet-161

86.6

DIS [63]

ResNet-101

86.8

APC-NET [37]

ResNet-101

87.1

HR Net V2 + OCR

HRNet-OCR

90.5

12.5 Conclusion Remarks In summary, it can be stated that there is a noteworthy enhancement in the performance of DL-based object detection and tracking methods. Over the past 4–5 years, there is a significant improvement of about 25–45% in mean IOU for multiple datasets. However, some articles lack reproducibility for several reasons such as, performance on nonstandard databases, or only perform on specific attributes but not on all, or does not have adequate experimental setups, most of the times, evaluate only subset of different object classes. Unfortunately, various publications did not deliver the source code for their implementations. Increasing of the popularity for deep learning networks, many scholars are implementing the reproducible frame works and giving open access for their works. Image segmentation was benefited highly from deep learning network, still there are lot of challenges to solve. In the future, some of the possible ways of researches like working on more datasets, integration of DNN and other segmentation methods, interpretable deep models, realtime models on different applications will support in further image segmentation approaches.

126

C. K. Kumar et al.

References 1. Szeliski, R.: Computer Vision: Algorithms and Applications. Springer (2010) 2. Otsu, N.: A threshold selection method from gray-level histograms. IEEE Trans. Syst. Man Cybern. 62–66 (1979) 3. Dhanachandra, N., Manglem, K., Chanu, Y.J.: Image segmentation using K-means clustering algorithm and subtractive clustering algorithm. Procedia Comput. Sci. 764–771 (2015) 4. Najman, L., Schmitt, M.: Watershed of a continuous function. Sig. Process. 99–112 (1994) 5. Kass, M., Witkin, A., Terzopoulos, D.: Snakes: active contour models. Int. J. Comput. Vis. 321–331 (1988) 6. Plath, N., Toussaint, M., Nakajima, S.: Multi-class image segmentation using conditional random fields and global classification. In: International Conference on Machine Learning, ACM, pp. 817–824 (2009) 7. Starck, J.L., Elad, M., Donoho, D.L.: Image decomposition via the combination of sparse representations and a variational approach. IEEE Trans. Image Process. 1570–1582 (2005) 8. Fukushima, K.: Neocognitron: a self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position. Biol. Cybern. 193–202 (1980) 9. Waibel, A., Hanazawa, T., Hinton, G., Shikano, K., Lang, K.J.: Phoneme recognition using time-delay neural networks. IEEE Trans. Acoust. Speech Signal Process. 328–339 (1989) 10. LeCun, Y., Bottou, L., Bengio, Y., Haffner, P., et al.: Gradient-based learning applied to document recognition. Proc. IEEE 2278–2324 (1998) 11. Krizhevsky, A., Sutskever, I., Hinton, G.E.: Imagenet classification with deep convolutional neural networks. Adv. Neural Inf. Process. Syst. 1097–1105 (2012) 12. He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778 (2016) 13. Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv 1409–1556 (2014) 14. Hochreiter, S., Schmidhuber, J.: Long short-term memory. Neural Comput. 1735–1780 (1997) 15. https://colah.github.io/posts/2015-08-Understanding-LSTMs/ 16. Badrinarayanan, V., Kendall, A., Cipolla, R.: SegNet: a deep convolutional encoder-decoder architecture for image segmentation. IEEE Trans. Pattern Anal. Mach. Intell. 2481–2495 (2017) 17. Radford, A., Metz, L., Chintala, S.: Unsupervised representation learning with deep convolutional generative adversarial networks. arXiv preprint arXiv 1511.06434 (2015) 18. Arjovsky, M., Chintala, S., Bottou, L.: Wasserstein GAN. arXiv preprint arXiv:1701.07875 (2017) 19. Mirza, M., Osindero, S.: Conditional generative adversarial nets. arXiv preprint arXiv:1411. 1784 (2014) 20. Long, J., Shelhamer, E., Darrell, T.: Fully convolutional networks for semantic segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 3431–3440 (2015) 21. Wang, G., Li, W., Ourselin, S., Vercauteren, T.: Automatic brain tumor segmentation using cascaded anisotropic convolutional neural networks. In: International MICCAI Brain lesion Workshop, pp. 178–190. Springer (2017) 22. Li, Y., Qi, H., Dai, J., Ji, X., Wei, Y.: Fully convolutional instance aware semantic segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 2359–2367 (2017) 23. Yuan, Y., Chao, M., Lo, Y.-C.: Automatic skin lesion segmentation using deep fully convolutional networks with Jaccard distance. IEEE Trans. Med. Imaging 1876–1886 (2017) 24. Liu, N., Li, H., Zhang, M., Liu, J., Sun, Z., Tan, T.: Accurate iris segmentation in noncooperative environments using fully convolutional networks. In: International Conference on Biometrics, pp. 1–8. IEEE (2016) 25. Liu, W., Rabinovich, A., Berg, A.C.: ParseNet: looking wider to see better. arXiv preprint arXiv:1506.04579 (2015) 26. Chen, L.-C., Papandreou, G., Kokkinos, I., Murphy, K., Yuille, A.L.: Semantic image segmentation with deep convolutional nets and fully connected CRFs. arXiv preprint arXiv:1412.7062 (2014)

12 Deep Network Architectures for Object Detection and Tracking: A Review

127

27. Schwing, A.G., Urtasun, R.: Fully connected deep structured networks. arXiv preprint arXiv: 1503.02351 (2015) 28. Lin, G., Shen, C., Van Den Hengel, A., Reid, I.: Efficient piecewise training of deep structured models for semantic segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 3194–3203 (2016) 29. Liu, Z., Li, X., Luo, P., Loy, C.-C., Tang, X.: Semantic image segmentation via deep parsing network. In: IEEE International Conference on Computer Vision, pp. 1377–1385 (2015) 30. Noh, H., Hong, S., Han, B.: Learning deconvolution network for semantic segmentation. In: IEEE International Conference on Computer Vision, pp. 1520–1528 (2015) 31. Fu, J., Liu, J., Wang, Y., Zhou, J., Wang, C., Lu, H.: Stacked deconvolutional network for semantic segmentation. IEEE Trans. Image Process. (2019) 32. Chaurasia, A., Culurciello, E.: LinkNet: Exploiting encoder representations for efficient semantic segmentation. In: IEEE International Conference on Visual Communications and Image Processing, pp. 1–4. IEEE (2017) 33. Xia, X., Kulis, B.: W-Net: a deep model for fully unsupervised image segmentation. arXiv preprint arXiv:1711.08506 (2017) 34. Lin, T.-Y., Doll´ar, P., Girshick, R., He, K., Hariharan, B., Belongie, S.: Feature pyramid networks for object detection. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 2117–2125 (2017) 35. Zhao, H., Shi, J., Qi, X., Wang, X., Jia, J.: Pyramid scene parsing network. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 2881–2890 (2017) 36. He, J., Deng, Z., Qiao, Y.: Dynamic multi-scale filters for semantic segmentation. In: IEEE International Conference on Computer Vision, pp. 3562–3572 (2019) 37. He, J., Deng, Z., Zhou, L., Wang, Y., Qiao, Y.: Adaptive pyramid context network for semantic segmentation. In: Conference on Computer Vision and Pattern Recognition, pp. 7519–7528 (2019) 38. He, K., Gkioxari, G., Doll´ar, P., Girshick, R.: Mask R-CNN. In: IEEE International Conference on Computer Vision, pp. 2961–2969 (2017) 39. Liu, S., Qi, L., Qin, H., Shi, J., Jia, J.: Path aggregation network for instance segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 8759–8768 (2018) 40. Dai, J., He, K., Sun, J.: Instance-aware semantic segmentation via multi-task network cascades. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 3150–3158 (2016) 41. Chen, L.-C., Hermans, A., Papandreou, G., Schroff, F., Wang, P., Adam, H.: Masklab: instance segmentation by refining object detection with semantic and direction features. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 4013–4022 (2018) 42. Pinheiro, P.O., Collobert, R., Doll´ar, P.: Learning to segment object candidates. Adv. Neural Inf. Process. Syst. 1990–1998 (2015) 43. Xie, E., Sun, P., Song, X., Wang, W., Liu, X., Liang, D., Shen, C., Luo, P.: PolarMask: single shot instance segmentation with polar representation. arXiv preprint arXiv:1909.13226 (2019) 44. Lee, Y., Park, J.: CenterMask: real-time anchor-free instance segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 13906–13915 (2020) 45. Chen, L.-C., Papandreou, G., Kokkinos, I., Murphy, K., Yuille, A.L.: DeepLab: semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFS. IEEE Trans. Pattern Anal. Mach. Intell. 834–848 (2017) 46. Wang, P., Chen, P., Yuan, Y., Liu, D., Huang, Z., Hou, X., Cottrell, G.: Understanding convolution for semantic segmentation. In: IEEE Winter Conference on Applications of Computer Vision, pp. 1451–1460 (2018) 47. Visin, F., Ciccone, M., Romero, A., Kastner, K., Cho, K., Bengio, Y., Matteucci, M., Courville, A.: ReSeg: a recurrent neural network-based model for semantic segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition Workshops, pp. 41–48 (2016) 48. Byeon, W., Breuel, T.M., Raue, F., Liwicki, M.: Scene labeling with LSTM recurrent neural networks. In: IEEE Conference on Comp Vision and Pattern Recognition, pp. 3547–3555 (2015)

128

C. K. Kumar et al.

49. Liang, X., Shen, X., Feng, J., Lin, L., Yan, S.: Semantic object parsing with graph LSTM. In: European Conference on Computer Vision. Springer, pp. 125–143 (2016) 50. Huang, Q., Xia, C., Wu, C., Li, S., Wang, Y., Song, Y., Kuo, C.-C.J.: Semantic segmentation with reverse attention. arXiv preprint arXiv:1707.06426 (2017) 51. Li, H., Xiong, P., An, J., Wang, L.: Pyramid attention network for semantic segmentation. arXiv preprint arXiv:1805.10180 (2018) 52. Fu, J., Liu, J., Tian, H., Li, Y., Bao, Y., Fang, Z., Lu, H.: Dual attention network for scene segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 3146– 3154 (2019) 53. Luc, P., Couprie, C., Chintala, S., Verbeek, J.: Semantic segmentation using adversarial networks. arXiv preprint arXiv:1611.08408 (2016) 54. Souly, N., Spampinato, C., Shah, M.: Semi supervised semantic segmentation using generative adversarial network. In: IEEE International Conference on Computer Vision, pp. 5688–5696 (2017) 55. Zheng, S., Jayasumana, S., Romera-Paredes, B., Vineet, V., Su, Z., Du, D., Huang, C., Torr, P.H.: Conditional random fields as recurrent neural networks. In: IEEE International Conference on Computer Vision, pp. 1529–1537 (2015) 56. Dai, J., He, K., Sun, J.: BoxSup: exploiting bounding boxes to supervise convolutional networks for semantic segmentation. In: IEEE International Conference on Computer Vision, pp. 1635– 1643 (2015) 57. Peng, C., Zhang, X., Yu, G., Luo, G., Sun, J.: Large kernel matters—improve semantic segmentation by global convolutional network. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 4353–4361 (2017) 58. Lin, G., Milan, A., Shen, C., Reid, I.: RefineNet: multi-path refinement networks for highresolution semantic segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1925–1934 (2017) 59. Wu, Z., Shen, C., Van Den Hengel, A.: Wider or deeper: revisiting the ResNet model for visual recognition. Pattern Recogn. 119–133 (2019) 60. Zhao, H., Zhang, Y., Liu, S., Shi, J., Change Loy, C., Lin, D., Jia, J.: PSANet: point-wise spatial attention network for scene parsing. In: European Conference on Computer Vision, pp. 267–283 (2018) 61. Zhang, H., Dana, K., Shi, J., Zhang, Z., Wang, X., Tyagi, A., Agrawal, A.: Context encoding for semantic segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 7151–7160 (2018) 62. Yu, C., Wang, J., Peng, C., Gao, C., Yu, G., Sang, N.: Learning a discriminative feature network for semantic segmentation. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1857–1866 (2018) 63. Luo, P., Wang, G., Lin, L., Wang, X.: Deep dual learning for semantic image segmentation. In: IEEE International Conference on Computer Vision, pp. 2718–2726 (2017) 64. Chen, L.-C., Zhu, Y., Papandreou, G., Schroff, F., Adam, H.: Encoder-decoder with atrous separable convolution for semantic image segmentation. In: European Conference on Computer Vision, pp. 801–818 (2018) 65. Zoph, B., Ghiasi, G., Lin, T.-Y., Cui, Y., Liu, H., Cubuk, E.D., Le, Q.V.: Rethinking pre-training and self-training. arXiv preprint arXiv:2006.06882 (2020) 66. Zhang, Z., Zhang, X., Peng, C., Xue, X., Sun, J.: ExFuse: enhancing feature fusion for semantic segmentation. In: European Conference on Computer Vision, pp. 269–284 (2018)

Chapter 13

Smart Vision of IoT: Semantic Approach of Data Analysis and Data Analytics in Agriculture 4.0 Smita Mane and Vaibhav E. Narawade

Abstract Nowadays, technology is one of the most important pillars in any smart development. Current systems are works on bounded efficiency, security, and productivity. The IoT is a ubiquitous intelligent territory for the more significant systems. With the utilization and applications of IoT in the traditional system, we develop more smarter, effective, and productive system. Our proposed research work purpose is to assess the importance of IoT in Agriculture 4.0 in order to construct smart and Agricultural 4.0 systems. Smart and intelligent IoT system concentrate on inefficiency of various factors like data integrity, power, time, security, and space which will be conclusively defend money and productivity. Ultimately the smart agricultural IoT system recognizes the problems occurs in traditional practices and justify the overall performance of system.

13.1 Introduction Nowadays due to rapid growth in the numerous technologies over and above, most of the world’s population have connection to the Web. Fourth industrial revaluation was initially well known in year 2015 which Industry 4.0 [1]. First, industrial revolution deals with hand production through machines use of steam power and water power. Second, industrial revolution is technological revolution used in railroad and telegraph networks. Third, revolution is digital revolution deals with super computer era. Very large-scale communication between machine to machine (M2M) and IoT is deeply integrated with each other, for more smarter, efficient, and secure system [2]. Industry 4.0 system bridges the gap into integrate environmental protection, control initiatives, and process safety [3]. Due to rapid growth of modern technology, various organizations are look forward for sustainability features in their various activities. With use of the Internet of things, various smart systems make Industry 4.0 efficient. S. Mane · V. E. Narawade (B) Ramrao Adik Institute of Technology, D Y Patil Deemed to be University, Mumbai, Maharashtra 400706, India e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_13

129

130

S. Mane and V. E. Narawade

Due to vast potential, Industry 4.0 has considered as leading technology framework used for integrating and extending in various sectors. Agricultural sector also adapting the conceptual view of the Industry 4.0. This concept sounds best at first, but it has multiple challenges that are to be overcome before implementing it. Due to an increase in competition among the various global technologies, it is important for researcher and developer to rethink about their operations and develop effective smart system [4]. In smart system driven program, primary point of convergence is utilization of huge amount of data impelled by IoT [5]. IoT driven data analytic combines new and old data that allows for a great deal of flexibility and decision-making capability to provide efficient system. Precision farming is used IoT in agriculture sector, so that farming is resulting in increased automation and lesser human intervention [6]. Smart system deals with capturing and analysis of various forms of data and establishes as data in real time. Smart system allows enterprises to be able to see and track assets in real time, advancements in the business/process, capitals, and products which optimize supply and demand.

13.2 Literature Review Initially, IoT issues and challenges with prospect to Agriculture 4.0 are studied, and based on this research, gap is found. Based on research gap, further survey is done on data analysis in Agriculture 4.0 with IoT smart system. Different fields emerged during in the fourth industrial revolution. In engineering, management system, manufacturing units, operations, logistical procedures, and agriculture Industry 4.0 will give significant flexibility and robustness, as well as the highest quality requirements. Industry 4.0 can be understood as the “smart system development in various domains which has the ability to influence the whole business in terms of products is designed, manufactured, and delivered” [3]. Agriculture 4.0 refers to the industry’s next major trends; due to the stronger emphasis on precision farming Agriculture 4.0 and use of data analytics drive greater economic efficiencies in the face of population growth and climate change. Testing of soil plays an very vital role nowadays in modern agriculture in order to optimize the productivity, protect the environment from excessive fertilizer usage, and save money and energy throughout production. In the year of 2018, the new term related to smart agriculture and precision farming was introduced by the World Government Summit. In India more than 40 percent population depend on agriculture sector for their livelihood. Day by day, this demand is increased, and in other hand, demand of precision farming is elevated, where all agricultural entities and stakeholders are connected together to form the agricultural 4.0. Smart farming allows minimizing wastage and increasing production and provides a clean and sustainable way of producing food. The different aspects of the Agriculture 4.0 include precision farming, smart irrigation, livestock monitoring, soil monitoring, crop monitoring, smart greenhouses.

13 Smart Vision of IoT: Semantic Approach of Data Analysis and Data …

131

13.2.1 Challenges and Issues Traditional farming is facing problem such as lack of communication, lack of realtime forecast, lack of education, lack of automation, and lack of reach [3]. In current scenario, agriculture and environmental monitoring have become easy due to presence of IoT sensors or devices. The available traditional techniques are like sowing, digging, and irrigation system. Due to long established methods of agricultural process, the Indian farmer faces many problems about productivity of agricultural product than others. It is due to unbalance feeding of fertilizer without knowing the specific requirement of nutrient to a crop [7]. The traditional soil testing is enhanced with use of technology, but still, the data analysis in the precision farming needed together to separate pieces of data. Secondly, real-time analytic refers to the collection, storage, processing, and distribution of data in real time so that analytical operations can be conducted on the dataset. Unlike in IoT regular data analysis applications, real-time applications must initiate fast actions that are usually bounded by specific data frames dictated by the targeted domain. As a result, maintaining data analysis solution should be approached with care, and it becomes even more important when a real-time approach is introduced in IoT. Descriptive analysis helps by developing key performance indicators (KPIs) and metrics such as return on investment (ROI). Analyzing data can optimize efficiency in precision agriculture. Elaborating the various parameters of product so that enrich the system with optimization can accomplish cutthroat world.

13.2.2 Comparative Analysis Agriculture 4.0 is described as having control over the degree, timing, and circumstances of sharing oneself with others (physically, behaviorally, or intellectually) [8]. Data analysis has become one of the most difficult tasks in research because to the rising use of computers, the Internet, and mobile phones. The goal is to design and implement a smart system in agriculture sector which fulfill the gap between technology and growth of farmers as well as economy. Allowing people to use these services effectively while optimizing the data from being exposed and inferred during the data analytic process. With the help of various technologies used in the Agriculture 4.0, sector comparative analysis is to be driven (Table 13.1).

13.3 Scope of Work The increased demand for food products has led to the concept of smart farming where all agricultural entities and stakeholders are connected together to form the Agricultural 4.0. Traditional farming is facing problem such as lack of communication, lack

Author

V.M. Abdul Hakkim et al. (2016)

Ojas Savale et al. (2016)

Shailaja Patil et al. (2016)

Amandeep et al. (2017)

Sr. No.

1

2

3

4

Table 13.1 Comparative analysis Technology used

Conclusion

ZigBee protocol, sensor

Image processing, GPS, CCTV

Wireless sensor network (WSN)

GPS-based vehicle and AVR-based greenhouse block in the agriculture sector will find better productivity

Automation in farming helps to improve the productivity with smart technologies

Improvement in the agriculture production with the help of IoT to achieve precision agriculture

GPS, GIS, sensor technology Precision agriculture is still only concept

Future scope

(continued)

Work on the precision farming with real-time data analytics

Development in the technology and smart phone application precision agriculture provides effective and smart solution to the farmers to grow with improving precision agriculture

Need to work on system which includes the smart system, which work on the action taken and monitoring intelligent system to take the decisions or actions according to the ruling status. So that the farmer’s participation with the system is reduced, less human work will be needed for monitoring

Need to comprises three phases’ exploration, analysis, and execution

132 S. Mane and V. E. Narawade

IoT sensor nodes including Arduino, Raspberry Pi, and Libelium Plug and Sense

Tomo Popovic et al. (2017)

Alessandro Massaro et al. (2018)

Ibrahim khider Eltahir et al. (2018)

Bhanu K.N et al. (2020

6

7

8

Concept and context view designed for precision farming. The stage has already been used for development of smart spraying and irrigation, valuation of the marine environment, and fish/mussel farm monitoring

Conclusion

Arduino UNO, YL-38 sensor, DHT11 sensor, DS18B20 sensor

DHT 11, ultrasonic sensors, Wi-Fi module

Analyze the soil parameters soil temperature, atmospheric temperature, and moisture

Solar supply and soil humidity analysis are done. The prediction helps to supply the right quantity of irrigation to the crops

ZigBee protocol, Data cloud work on weather and GSM/GPRS, DSS Algorithm field station with evaluation of DSS algorithm

Technology used

Author

Sr. No.

5

Table 13.1 (continued)

Correlation between the parameter and predict the fertility of the soil by considering soil fertility parameters

Data analytics must enhance in the prediction of outcomes

New or advanced DSS irrigation algorithm with integrating fertirrigation

Need work on MQTT protocol integrated with R and Python for comprehensive data analytics. Development of matrices for betterment of scalability, reliability, and performance

Future scope

13 Smart Vision of IoT: Semantic Approach of Data Analysis and Data … 133

134

S. Mane and V. E. Narawade

of real-time forecast, lack of education, lack of automation, and lack of reach. In current scenario, agriculture and environmental monitoring have become easy due to presence of IoT sensors or devices. The traditional soil testing is enhanced with use of technology, but still, the data analysis in the precision farming needed together to separate pieces of data. Secondly, real-time analytic refers to the collection, storage, processing, and distribution of data in real time so that analytical operations can be conducted on the dataset. As a result, maintaining data analysis solution should be approached with care, and it becomes even more important when a real-time approach is introduced in IoT. Descriptive analysis helps by developing key performance indicators (KPIs) and metrics such as return on investment (ROI). Analyzing data can optimize efficiency in precision agriculture. Elaborating the various parameters of product so that enrich the system with optimization can accomplish cutthroat world.

13.4 Methodology The current available soil testing terminology is collection of soil as per the laboratory given technique. After collecting the soil samples, it is deputed to lab for examination. This is very time consuming process. Agriculture 4.0 enabling technologies are still in baby crawling or infancy stage, and firms are far away from enjoying its complete benefits. The perception level of agriculture sector on smart IoT system is very low. Agriculture 4.0 system is not well adopted due to various gap occurs in this sector. The growth of Agricultural 4.0 depends upon the progress in academic research. Unavailability of strong empirical data analysis and data analytics is the main reason for less adoption IoT in Agriculture 4.0 smart system. Due to a lack of implementation details and a large amount of fund requirements, many sectors are still not sure with the use of these new technologies. Therefore, it is a need of time to a research into an Agricultural 4.0 (Fig. 13.1).

Fig. 13.1 Workflow of the proposed system

13 Smart Vision of IoT: Semantic Approach of Data Analysis and Data …

135

13.5 Result and Discussion We are work on the 4 Vs, such as volume which is scale of data, variety is different forms of data, velocity is analysis of streaming data, and veracity is uncertainty of data. Our proposed research work is wounded within three parts, that is, one is sensor data real-time data collection and processing the data with the help of data optimization, and third is visualize the data with remote data access. With the help of these three phases, we can form the unstructured data in to the structured data in soil testing. With the help of this, we use the agriculture services effectively while optimizing the data from being exposed and inferred during the data analytic process. The science of analyzing data to develop conclusions is known as data analytics. With tremendous growth of digital technology, plenty or larger number of data including their identity, specification, environment, and location connected via IoT. Data administration and analysis will be complicated by IoT data. For the data analytics, we follow the various steps, first are examining the data, pre-processing the data set, data cleaning to avoid the data redundancy, and finally explore the data and model it. Graph theory addresses object interactions through geometric structures (graphs). It was designed in the 1700s by the famed Swiss mathematician Leonard Euler to solve the Konigsberg Bridge Problem. In our proposed research work, we work on the Kalyan district data set of soil testing. Parameters such as Ph, EC, OC, N, K, S, Zn, Mn, and C are the parameters on which the soil testing is to perform. In agriculture sector to increase the productivity, soil is the important parameter on which nutrient deficiency is to be evaluated. With the help of soil testing, various parameters are too analyzed for the optimization of crop production. With the help of sensor node, we can collect the real-time data, which play a key role in precision agriculture system. These nodes can make the system more realistic by gathering real-time data from soil testing areas to improve the precision of the agriculture system. By incorporating data analytics and machine learning into to the agriculture sector, farming becomes more economical. To apply the sensor in the Agriculture 4.0, smart soil testing system process, domain, and information model specification are to be required, which is to be incorporated with the help of above data examination and pre- processing shown in Figs. 13.2, 13.3, and 13.4. For the selection of sensor nodes parameter range, least count/sensitivity and deviations are to be framed.

13.6 Conclusion Various smart applications for farmers are being invented in precision agriculture to keep them up to date on the status of increase their productivity. Every agrarian activity is built on the foundation of soil. There is no such thing as a crop without soil. The stomach of plant life is thought to be the soil. So analyzing the state of the soil is the first and most important step toward the optimum agricultural technique. The proposed system for smart agriculture system is usually formed by the various

136

S. Mane and V. E. Narawade

Fig. 13.2 pH, Ec, and OC analysis graph

Fig. 13.3 Fe, Mn, and Cu analysis graph

phases which is shown in Fig. 13.1. For the real-time data collection of specified soil parameter, we required various sensor nodes, for e.g., a soil moisture sensor measures how wet the soil is, whereas a soil nutrient sensor measures how fertile the soil is. With the help of this phase, we collect the real-time categorical data. Depending on the necessity, we can design the data visualization and data storing management by local or cloud enhanced computing. For the selection of the optimal sensor node, data examination and pre-processing are to be performed, and sensor node selection parameters are designed. Table 13.2 is specified the various twelve parameters with its range, which will be help helpful for the sensor node selection and data analysis. Data analysis techniques are used to determine the state of the soil parameters. These

13 Smart Vision of IoT: Semantic Approach of Data Analysis and Data …

137

Fig. 13.4 N, P, K, and S analysis graph

values are used to selection of threshold value of sensor nodes. We can examine the physical, biological, and chemical characteristics of the soil by doing soil testing. Farmers can make knowledgeable decisions concerning their fields based on this information. Precision farming is primarily motivated by the desire to produce more from the limited amount of productive agricultural land available. Table 13.2 Sensor node selection parameters Sr. No

Parameter

Range

1

pH

0.08 deviations ± to 4%

2

EC

Deviations ± 10% temperature correction is now introduced

3

OC

Range 0.2–1.5%

4

N

Indirect, based on organic C

5

P

0.9 kg/ha, deviation ± 5%

6

K

The detection is 1.29 kg/ha for alluvial soil, 2.19 kg/ha for black soil, and 1.39 kg ha for red soil

7

S

Detection 1 mg/kg. Deviation within ± 10%

8

B

Detection: 0.09 mg/kg. Critical limit 0.5 mg/kg

9

Fe

Detection: 0.11 mg/kg. Critical limit 4.5 mg/kg

10

Mn

Detection: 0.45 mg/kg. Critical limit 2.0 mg/kg

11

Cu

Detection: 0.09 mg/kg. Critical limit 0.2 mg/kg

12

Zn

Detection: 0.18 mg/kg (in some cases 10–15%)

138

S. Mane and V. E. Narawade

13.7 Future Scope With the help of above result and analysis, we are obtained sensor node selection parameter with threshold values. For the future work, we focus on service specification and IoT level specification.

References 1. Sung, J.: The fourth industrial revolution and precision agriculture. In: Automation in Agriculture Securing Food Supplies for Future Generations, pp. 1–15 (2018) 2. Atzori, L., Iera, A., Morabito, G.: Understanding the 5 internet of things definition, potentials, and societal role of a fast evolving paradigm. Ad Hoc Netw. 56, 122–140 (2017) 3. Misaki, E., Apiola, M., Gaiani, S., Tedre, M.: Challenges facing sub-Saharan small-scale farmers in accessing farming information through mobile phones: a systematic literature review. Electronic J. Inf. Syst. Developing Countries 84(4), e12034 (2018) 4. Ryan, P.J., Watson, R.B.: Research challenges for the internet of things: what role can or play. Systems 5, 24 (2017). https://doi.org/10.3390/systems5010024 5. El-Gayar, O., Ofori, M.: Disrupting Agriculture: The Status and Prospects for AI and Big Data in Smart Agriculture (2020) 6. Saidu, A., Clarkson, A.M., Adamu, S.H., Mohammed, M., Jibo, I.: Application of ICT in agriculture: opportunities and challenges in developing countries. Int. J. Comput. Sci. Math. Theor. (2017) 7. Keller, M., Rosenberg, M., Brettel, M., Friederichsen, N.: How virtualization, decentrazliation and network building change the manufacturing land scape: an industry 4.0 perspective. Int. J. Mech. Aerospace Ind. Mechatron. Manuf. Eng. 8, 37–44 (2014) 8. Connolly, A.J., Phillips-Connolly, K.: Can agribusiness feed bil- lion new people and save the planet? A glimpse into the future. Int. Food Agribusiness Manage. Rev. 15, 14 (2012)

Chapter 14

Hybrid Boost Converter Integrated Seven-Level MLI Fed PMSM Drive with Closed-Loop Speed Control J. A. Ganeswari and R. Kiranmayi

Abstract DC–DC converters are most widely used in distributed generation. The main objective of boost converter is to enhance the output voltage. Lot of boost converter topologies have been proposed till now. In this paper, hybrid boost converter is proposed and compared with the conventional type. This paper presents the analysis of seven-level diode clamped multi-level inverter fed PMSM which is driven from pulse generator employing level-shifted (multi) carrier PWM pattern. Reference current signal is generated from closed-loop control of PMSM. Closed-loop speed control is presented in detail. The detailed analysis of the proposed converter circuit is carried out by simulating the converter circuit using Matlab/simulink software.

14.1 Introduction The socioeconomic development of a country depends on the electricity, and it is basic requirement of infrastructural. There are mainly two challenges in the developing countries because of increased energy demand such as integrating new power plants and expansion of the transmission and distribution system. These two issues can be solved by distributed generation (DG) [1–3]. The distributed generation can be installed at the load centers or at distribution level by using renewable energy sources. The renewable energy sources RES [4–6] are solar, wind, and fuel cell. The output voltage from these renewable energy sources is very low in magnitude and DC in nature. In order to connect these renewable energy sources to utility grid, the low voltage DC must be stepped up to required voltage level. For this purpose, a DC–DC converter is needed. There are lot of DC–DC converter topologies [4–6] which have been proposed, and the basic converter is the conventional boost converter. The step up operation in the boost converter is based on the inductor and the duty ratio. To J. A. Ganeswari (B) EEE Department, JNTUA, Anantapuramu, India e-mail: [email protected] R. Kiranmayi Foreign Affairs & Alumni Matters, JNTUA CEA, Anantapuramu, India © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_14

139

140

J. A. Ganeswari and R. Kiranmayi

get the high step up voltage, the duty ratio can be increased, or large inductance must be used. But, the losses will be increased which are associated with power switch, inductor, and capacitor. Therefore, Hybrid boost converter is proposed. In hybrid boost converter, two inductors are used along with two power semiconducting switches. These inductors and switches are connected in Z shape. This hybrid boost converter gives the high voltage gain. PMSM motors are generally fed from sinusoidal source. If PMSM is to be fed from a DC source, inverter (DC–AC) does the job of supplying sinusoidal excitation to phase windings of PMSM. Conventional inverters (two-level or square wave) suffer from high dv/dt across switch resulting in high switching losses. High switching losses reduce the efficiency of the converter. Nowadays, multi-level inverters became the prior pick in many of the industries for high power high voltage applications. Multi-level inverters are able to generate high voltage with lower rated devices. Multi-level inverter generates leveled (stepped) output, and as the number of level increases, better output voltage waveform is obtained. Diode clamped multi-level inverters are one among multi-level topologies and use diodes as clamping elements. The paper presents the analysis of seven-level diode clamped multi-level inverter fed PMSM. Seven-level diode clamped inverter is driven from pulse generator employing level-shifted (multi) carrier PWM pattern.

14.2 Hybrid Boost Converter (a) Main circuit The hybrid boost converter is shown in Fig. 14.1. It has two switches with two inductors. In order to increase the voltage gain, additional inductors and switches have been used. These are connected in the Z shape. (b) Switching operation Figure 14.2a, b shows the switching operation of the hybrid boost converter during switches ON and OFF, respectively. When switch is ON, the two inductors get charged by the voltage source, and load is supplied by output capacitor. During OFF condition, the charged inductors are discharge to the load. L1

Fig. 14.1 Hybrid boost converter Vin

S2

S1

L2

C

R

14 Hybrid Boost Converter Integrated Seven-Level MLI Fed PMSM Drive …

141

L1

Vin

L1

L2

C

R

C

Vin

R

L2

(a) During switch ON

(b) During switch OFF

Fig. 14.2 Hybrid boost converter switching operation. a During switch ON, b during switch OFF

(c) Voltage gain derivation The derivation for voltage gain formula is given below.

 Vin ∗ DTs +

 VLavg = 0

 Vin −V0  2

(1 − D)Ts

Ts Vgainproposed =

=0

1+ D V0 = Vin 1− D

14.3 Seven-Level Diode Clamped Inverter Fed PMSM Seven-level DCMLI is shown in Fig. 14.4. Seven-level DCMLI consists of twelve power switches, ten diodes, and six capacitors. Power switches are generally IGBTs. The main DC voltage source is split equally across all the six capacitors to give leveled output waveform. Table 14.1 illustrates the switching states of twelve power switches in seven-level DCMLI to give out seven-level stepped output waveform. Sequential switching of power switches in phase yields seven-levels of output as shown in (Figs. 14.3, 14.4 and 14.5).

142

J. A. Ganeswari and R. Kiranmayi

Table 14.1 Comparison between converters

Boost converter

Hybrid boost converter

Input current ripple

4A

70 A

Output voltage ripple

8V

7V

Inductor current

100 A

63 A

Fig. 14.3 Control strategy for PMSM drive Fig. 14.4 Seven-level output of DCMLI

14 Hybrid Boost Converter Integrated Seven-Level MLI Fed PMSM Drive …

143

Fig. 14.5 Control of PMSM drive

The control strategy involves simple arithmetic calculations and blocks to control the speed of PMSM. Control algorithm accepts mechanical feedback from the PMSM to generate reference currents for the pulse generation. Actual speed of the PMSM is sensed and is compared to the reference speed signal. The error in speed is fed to PI controller. PI controller reduces the error signal and generates direct axis component of current signal. Quadrature component of current is considered to be zero. Both the quadrature and direct axes components are processed to inverse Park’s transformation to generate reference current signal in “abc” co-ordinate system.

14.4 Simulation Results The switch voltage and switch current are shown in Fig. 14.12. It is seen that when the switch is ON, the voltage across it is zero, and when the switch is OFF, the voltage across it is 400 V. At the same time, when the switch is ON, the current passes through it is 100 A, and when the switch is OFF, no current passes through it. (a) Hybrid Boost Converter The input voltage and input current of hybrid boost converter are shown in Fig. 14.6. The magnitude of input voltage is 100 V constant. The input current has the ripples between 60 and 130 A. Figure 14.7. The output voltage has the magnitude of 398 V with a ripple between 395 and 402 V. The output current has the magnitude of 25 A with ripple between 24.7 and 25.2 A (Figs. 14.8 and 14.9) Case 1: PMSM with fixed speed. Figure 14.10 gives the speed curve of PMSM running with fixed speed condition. PMSM is set to run at constant 3000 RPM, and actual speed follows the reference speed (Figs. 14.11, 14.12, 14.13, 14.14, 14.15 and 14.16). Case 2: PMSM with variable speed. Figure 14.17a PMSM is set to run at constant 2000 RPM initially, and actual speed follows the reference speed. After 0.25 s time, set speed changes to 3500 RPM, and

144

J. A. Ganeswari and R. Kiranmayi

(a)

(b)

Fig. 14.6 a Input voltage, b input current

(a)

(b)

Fig. 14.7 a Output voltage, b output current

(a)

(b)

Fig. 14.8 a Switch voltage of 250 V, b switch current of 60 A

(a) Fig. 14.9 a Diode voltage of 500 V, b diode current of 60 A

(b)

14 Hybrid Boost Converter Integrated Seven-Level MLI Fed PMSM Drive …

Fig. 14.10 Speed of PMSM at fixed speed 3000 RPM

Fig. 14.11 Stator currents of PMSM at fixed speed 3000 RPM

Fig. 14.12 Load torque of PMSM at fixed speed 3000 RPM

Fig. 14.13 Torque generated by PMSM at fixed speed 3000 RPM

145

146

J. A. Ganeswari and R. Kiranmayi

(a)

(b)

Fig. 14.14 a THD in stator currents, b THD in phase voltage of DCMLI harmonic distortion in stator currents by 1.27% with respect to fundamental, and Fig. 14.14 shows phase voltage of DCMLI which is distorted by 23.47%

Fig. 14.15 Line voltage of 50 V DCMLI

actual speed follows reference value, and Fig. 14.17b shows the load torque of 5 Nm which is impressed on PMSM. Figure 14.18a shows the harmonic distortion in stator currents by 2.36% with respect to fundamental, and Fig. 14.18b shows the harmonic distortion in phase voltage of DCMLI by 31.35%. Figure 14.19a, b shows the output phase voltage of diode clamped inverter, and it generates only half the amount of DC voltage source. DC voltage fed to DCMLI is 500 V, and phase voltage obtained is with peak 250 V. After 0.25 s, the frequency of phase voltage increases as shown. THD FFT window of line voltage is shown in Fig. 14.20. Line voltage is distorted by 23.38% (Fig. 14.21).

14 Hybrid Boost Converter Integrated Seven-Level MLI Fed PMSM Drive …

147

Fig. 14.16 Line voltage is distorted by 20.80%

(a)

(b)

Fig. 14.17 a Speed of PMSM, b load torque of PMSM

Table 14.2 shows the comparison of harmonic distortion analysis with different running conditions of PMSM.

148

J. A. Ganeswari and R. Kiranmayi

(a)

(b)

Fig. 14.18 a THD in stator currents, b THD in phase voltage of DCMLI

(a)

(b)

Fig. 14.19 a Output phase voltage of DCMLI, b line voltage of DCMLI

14.5 Conclusion In conclusion, it is firmly proved that the hybrid boost converter has fewer ripples in input current and output voltage. Also, the current carried by the inductor is less as compared to hybrid boost converter. In this paper, the simulation analysis is done for conventional boost converter and hybrid boost converter. In case of hybrid boost converter, the input current ripple is 70 amps, the output voltage ripple is 7 V, and the current carried by the inductor is 63 amps. Characteristics of seven-level DCMLI are shown. PMSM characteristics are shown with fixed and variable speed operating conditions. The control algorithm presented effectively controls the speed of PMSM in fixed and variable speed conditions. In-phase LSCPWM, multi-carrier PWM drives power switches of DCMLI. THD analysis is compared for different running conditions of PMSM. Stator current is distorted by very less quantity, which implies sinusoidal excitation is supplied to phase windings of PMSM.

14 Hybrid Boost Converter Integrated Seven-Level MLI Fed PMSM Drive …

Fig. 14.20 THD of line voltage

Fig. 14.21 Stator currents of PMSM

149

150 Table 14.2 Comparison of THD

J. A. Ganeswari and R. Kiranmayi THD

Fixed speed mode (%)

Variable speed mode (%)

Stator current of PMSM

1.27

2.36

Phase voltage of inverter

23.47

31.35

Line voltage of inverter

20.80

23.38

References 1. Verdelho, M. I., Prata, R., Carvalho, P., Machado, J.: Impact of PV distributed generation on EDP distribuição LV grid losses. In: CIRED—Open Access Proceedings Journal, vol. 2017, no. 1, pp. 2342–2345 (2017) 2. Shah, P., Hussain, I., Singh, B.: Real-time implementation of optimal operation of single-stage grid interfaced PV system under weak grid conditions. In: IET Generation, Transmission & Distribution, vol. 12, no. 7, pp. 1631–1643 (2018) 3. Ranjan, R.: Solid state relay based inrush current limiter with short circuit and under voltage protection for dc-dc converters. In: 2017 IEEE International Conference on Consumer Electronics-Asia (ICCE-Asia), Bangalore, pp. 47–51 (2017) 4. Samaranayake, L., Chin, Y.K., Alahakoon, U.S.K.: Distributed control of permanent magnet synchronous motor drive systems. In: The Fifth International Conference on Power Electronics and Drive Systems, PEDS 2003, vol. 1, pp. 710–715 (2003) 5. Sree Lakshmi, G., Kamakshaiah, S., Das, T.R.: Closed loop PI control of PMSM for hybrid electric vehicle using three level diode clamped inverter for optimal efficiency. In: 2013 International Conference on Energy Efficient Technologies for Sustainability (ICEETS), Nagercoil, pp. 754–759 (2013) 6. Sekar, R., Suresh, D.S., Naganagouda, H.: A review on power electronic converters suitable for renewable energy sources. In: 2017 (ICEECCOT), Mysuru, pp. 501–506 (2017)

Chapter 15

A Braille Learning Device for the Visually Impaired Pranav Sakre , Sanika More , and Shruti Dodani

Abstract India stands home to twenty percent of the worlds visually impaired. We have observed a tremendous development in technology which has made our lives simpler and more convenient. Somehow, the visually impaired part of society remains disregarded. The purpose of this project is to create a single tool to facilitate the visually impaired with textual content access. The same old QWERTY keyboard is inaccessible to the blind, as it, in most cases, calls for finding keys’ location by using their fingers. Most keyboards available on the market for the visually impaired consist of the keyboard with braille alphanumerical versions etched at the keys. This paper presents a type of keyboard for the visually impaired which would not only help them learn braille but also serve as a concise, portable, and intuitive braille keyboard, thereby replacing the conventional keyboard. The proposed keyboard replaces the QWERTY keyboard with just 12 keys, out of which 6 are used for acquiring characters according to the braille system. The other keys represent unique keys like space, backspace, caps lock, speak, and cursor navigation keys. With an audio output, the visually impaired can hear the alpha numerals or words that they type.

15.1 Introduction Visual impairment is a condition that limits the visible functionality of a person to such an extent that additional assistance is needed. Many people believe that because of advanced assistive technology like speech tools, the visually impaired can do the same things as those with perfect vision. However, the ability to read is a very important part of one’s cognitive development. Braille is a crucial tool that enables the visually impaired to become literate and keep up with the rest of the world. P. Sakre (B) · S. More · S. Dodani Department of Biomedical Engineering, Dwarkadas J. Sanghvi College of Engineering, Mumbai, India e-mail: [email protected] S. Dodani e-mail: [email protected] © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2023 A. R. Manchuri et al. (eds.), Intelligent Manufacturing and Energy Sustainability, Smart Innovation, Systems and Technologies 334, https://doi.org/10.1007/978-981-19-8497-6_15

151

152

P. Sakre et al.

Table 15.1 Definitions of blindness and visual impairment [2] Type

Definition

Severe visual impairment (SVI)

Vision in the better eye with available correction is between 20/200–20/400

Moderate visual impairment (MVI)

Vision in the better eye with available correction is between 20/70–20/160

Early visual impairment (EVI)

Visual acuity is worse than 6/12–6/18

Moderate severe visual impairment (MSVI)

Vision in the better eye with available correction is less than 6/18–3/60

Visual impairment (VI)

Visual loss with visual acuity less than either 20/40 or 20/60

Functional low vision

Impairment of visual functioning even after treatment, having visual acuity