Full Ebook of Advances in Data Science and Computing Technology Suman Ghosal Online PDF All Chapter

Advances in Data Science and
Computing Technology Suman Ghosal

Visit to download the full and correct content document:
https://ebookmeta.com/product/advances-in-data-science-and-computing-technology-
suman-ghosal/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Advances in Data Science and Computing Technology:

Methodology and Applications 1st Edition Suman Ghosal
(Editor)
https://ebookmeta.com/product/advances-in-data-science-and-
computing-technology-methodology-and-applications-1st-edition-
suman-ghosal-editor/
Advances in Digital Science ICADS 2021 Advances in

Intelligent Systems and Computing 1352 Tatiana
Antipova (Editor)
https://ebookmeta.com/product/advances-in-digital-science-
icads-2021-advances-in-intelligent-systems-and-
computing-1352-tatiana-antipova-editor/
Advances in Data Science 1st Edition Ilke Demir
https://ebookmeta.com/product/advances-in-data-science-1st-
edition-ilke-demir/
Advanced Soft Computing Techniques in Data Science, IoT

and Cloud Computing Sujata Dash (Editor)
https://ebookmeta.com/product/advanced-soft-computing-techniques-
in-data-science-iot-and-cloud-computing-sujata-dash-editor/
Recent Advances in Artificial Intelligence and Data
Engineering: Select Proceedings of AIDE 2020 (Advances
in Intelligent Systems and Computing, 1386)
https://ebookmeta.com/product/recent-advances-in-artificial-
intelligence-and-data-engineering-select-proceedings-of-
aide-2020-advances-in-intelligent-systems-and-computing-1386/
Sustainable Production and Applications of Waterborne

Polyurethanes Advances in Science Technology Innovation
https://ebookmeta.com/product/sustainable-production-and-
applications-of-waterborne-polyurethanes-advances-in-science-
technology-innovation/
Introduction to Quantum Computing (River Publishers

Series in Rapids in Computing and Information Science
and Technology) 1st Edition Ahmed Banafa
https://ebookmeta.com/product/introduction-to-quantum-computing-
river-publishers-series-in-rapids-in-computing-and-information-
science-and-technology-1st-edition-ahmed-banafa/
The Farinograph Handbook : Advances in Technology,

Science, and Applications, 4th Edition Jayne E. Bock
https://ebookmeta.com/product/the-farinograph-handbook-advances-
in-technology-science-and-applications-4th-edition-jayne-e-bock/
Frontiers of Data and Knowledge Management for

Convergence of ICT, Healthcare, and Telecommunication
Services (EAI/Springer Innovations in Communication and
Computing) Suman Paul (Editor)
https://ebookmeta.com/product/frontiers-of-data-and-knowledge-
management-for-convergence-of-ict-healthcare-and-
telecommunication-services-eai-springer-innovations-in-
ADVANCES IN DATA SCIENCE AND
COMPUTING TECHNOLOGY
Methodology and Applications
AAP Research Notes on Optimization & Decision-Making Theories
ADVANCES IN DATA SCIENCE AND

COMPUTING TECHNOLOGY
Edited by
Suman Ghosal, PhD
Amitava Choudhury, PhD
Vikram Kumar Saxena, MTech
Arindam Biswas, PhD
Prasenjit Chatterjee, PhD
First edition published 2023
Apple Academic Press Inc. CRC Press
1265 Goldenrod Circle, NE, 6000 Broken Sound Parkway NW,
Palm Bay, FL 32905 USA Suite 300, Boca Raton, FL 33487-2742 USA
760 Laurentian Drive, Unit 19, 4 Park Square, Milton Park,
Burlington, ON L7N 0A4, CANADA Abingdon, Oxon, OX14 4RN UK
© 2023 by Apple Academic Press, Inc.

Apple Academic Press exclusively co-publishes with CRC Press, an imprint of Taylor & Francis Group, LLC
Reasonable efforts have been made to publish reliable data and information, but the authors, editors, and publisher cannot assume
responsibility for the validity of all materials or the consequences of their use. The authors, editors, and publishers have attempted
to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to
publish in this form has not been obtained. If any copyright material has not been acknowledged, please write and let us know so
we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any
form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and
recording, or in any information storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.com or contact the Copyright
Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. For works that are not available on CCC
please contact mpkbookspermissions@tandf.co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks and are used only for identification
and explanation without intent to infringe.
Library and Archives Canada Cataloguing in Publication

Title: Advances in data science and computing technology : methodology and applications / edited by Suman Ghosal, PhD,
Amitava Choudhury, PhD, Vikram Kumar Saxena, MTech, Arindam Biswas, PhD, Prasenjit Chatterjee, PhD.
Names: Ghosal, Suman, editor. | Choudhury, Amitava, editor. | Saxena, Vikram Kumar, editor. | Biswas, Arindam, editor. |
Chatterjee, Prasenjit, 1982- editor.
Description: First edition. | Series statement: AAP research notes on optimization and decision-making theories | Includes
bibliographical references and index.
Identifiers: Canadiana (print) 20220220735 | Canadiana (ebook) 20220220808 | ISBN 9781774639979 (hardcover) |
ISBN 9781774639986 (softcover) | ISBN 9781003277071 (ebook)
Subjects: LCSH: Computer systems. | LCSH: Electronic data processing.
Classification: LCC QA76 .A38 2023 | DDC 004—dc23
Library of Congress Cataloging-in-Publication Data
CIP data on file with US Library of Congress
ISBN: 978-1-77463-997-9 (hbk)

ISBN: 978-1-77463-998-6 (pbk)
ISBN: 978-1-00327-707-1 (ebk)
AAP RESEARCH NOTES ON
OPTIMIZATION AND
DECISION-MAKING THEORIES
SERIES EDITORS:
Dr. Prasenjit Chatterjee
Department of Mechanical Engineering, MCKV Institute of Engineering,
Howrah, West Bengal, India
E-Mail: dr.prasenjitchatterjee6@gmail.com / prasenjit2007@gmail.com
Dr. Dragan Pamucar
University of Defence, Military academy, Department of Logistics,
Belgrade, Serbia; E-Mail: dpamucar@gmail.com
Dr. Morteza Yazdani
Department of Business & Management, Universidad Loyola Andalucia,
Seville, Spain; E-Mail: morteza_yazdani21@yahoo.com
Dr. Anjali Awasthi
Associate Professor and Graduate Program Director (M.Eng.), Concordia
Institute for Information Systems Engineering, Concordia, Canada
E-Mail: anjali.awasthi@concordia.ca
Most real-world search and optimization problems naturally involve multiple

criteria as objectives. Different solutions may produce trade-offs (conflicting
scenarios) among different objectives. A solution that is better with respect
to one objective may be a compromising one for other objectives. This
compels one to choose a solution that is optimal with respect to only one
objective. Due to such constraints, multi-objective optimization problems
(MOPs) are difficult to solve since the objectives usually conflict with each
other. It is usually hard to find an optimal solution that satisfies all objectives
from the mathematical point of view. In addition, it is quite common
that the criteria of real-world MOPs encompass uncertain information,
which becomes quite a challenging task for a decision maker to select the
criteria. Also, the complexities involved in designing mathematical models
increase. Considering, planning, and appropriate decision-making require
vi AAP Research Notes on Optimization and Decision-Making Theories
the use of analytical methods that examine trade-offs; consider multiple

scientific, political, economic, ecological, and social dimensions; and reduce
possible conflicts in an optimizing framework. Among all these, real-world
multi-criteria decision-making (MCDM) problems related to engineering
optimizations are categorically important and are quite often encountered
with a wide range of applicability.
MCDM problems are basically a fundamental issue in various fields,
including applied mathematics, computer science, engineering, management,
and operations research. MCDM models provide a useful way for modeling
various real-world problems and are extensively used in many different
types of systems, including, but not limited to, communications, mechanics,
electronics, manufacturing, business management, logistics, supply chain,
energy, urban development, waste management, and so forth.
In the aforementioned cases, modeling of multiple criteria problems
often becomes more complex if the associated parameters are uncertain and
imprecise in nature. Impreciseness or uncertainty exists within the param
eters due to imperfect knowledge of information, measurement uncertainty,
sampling uncertainty, mathematical modeling uncertainty, etc. Theories like
probability theory, fuzzy set theory, type-2 fuzzy set theory, rough set, grey
theory, neutrosophic uncertainty theory available in the existing literature
deal with such uncertainties. Nevertheless, the uncertain multi-criteria
characteristics in such problems are not explored in depth, and a lot can be
achieved in this direction. Hence, different mathematical models of real-life
multi-criteria optimization problems can be developed on various uncer
tain frameworks with special emphasis on sustainability, manufacturing,
communications, biomedical, electronics, materials, energy, agriculture,
environmental engineering, strategic management, flood risk management,
supply chain, waste management, transportations, economics, and industrial
engineering problems, to name a few.
Coverage & Approach:
The primary endeavor of this series is to introduce and explore contemporary
research developments in a variety of rapidly growing decision-making
areas. The volumes will deal with the following topics:
• Crisp MCDM models
• Rough set theory in MCDM
• Fuzzy MCDM
• Neutrosophic MCDM models
• Grey set theory
AAP Research Notes on Optimization and Decision-Making Theories vii
• Mathematical programming in MCDM

• Big data in MCDM
• Soft computing techniques
• Modelling in engineering applications
• Modeling in economic issues
• Waste management
• Agricultural practice
• Material selection
• Renewable energy planning
• Industry 4.0
• Sustainability
• Supply chain management
• Environmental policies
• Manufacturing processes planning
• Transportation and logistics
• Strategic management
• Natural resource management
• Biomedical applications
• Future studies and technology foresight
• MCDM in governance and planning
• MCDM and social issues
• MCDM in flood risk management
• New trends in multi-criteria evaluation
• Multi-criteria analysis in circular economy
• Multi-criteria evaluation for urban and regional planning
• Integrated MCDM approaches for modeling relevant applications
and real-life problems
Types of volumes:
This series reports on current trends and advances in optimization and
decision-making theories in a wider range of domains for academic and
research institutes along with industrial organizations. The series will cover
the following types of volumes:
• Authored volumes
• Edited volumes
• Conference proceedings
• Short research (thesis-based) books
• Monographs
viii AAP Research Notes on Optimization and Decision-Making Theories
Features of the volumes will include recent trends, model extensions,

developments, real-time examples, case studies, and applications. The
volumes aim to serve as valuable resources for undergraduate, postgraduate
and doctoral students, as well as for researchers and professionals working
in a wider range of areas.
CURRENT & FORTHCOMING BOOKS IN THE SERIES

Multi-Criteria Decision-Making Techniques in Waste Management:
A Case Study of India
Suchismita Satapathy, Debesh Mishra, and Prasanjit Chatterje
Applications of Artificial Intelligence in Business and Finance:
Modern Trends
Editors: Vikas Garg, Shalini Aggarwal, Pooja Tiwari, and Prasenjit Chatterjee
Advances in Data Science and Computing Technology:
Editors: Suman Ghosal, Amitava Choudhury, Vikram Kr. Saxena, Arindam Biswas,
and Prasenjit Chatterjee
Machine Learning Algorithms in Security Analytics: Applications,
Principles, and Practices
Editors: Karan Singh, Latha Banda, and Manisha Manjul
Sustainable Development: A Geospatial and Statistical Perspective
Editors: Qazi Mazhar Ali, Rizwan Ahmad, Irfan Ali, Prasenjit Chatterjee, and
Harjit Pal Singh
Decision Analysis and Optimization Models for Sustainable
Development Goals 2030
Irfan Ali, Umar Muhammad Modibbo, and Bui Thanh Hung
Smart Sensors, Actuators, and Decision Support Systems for Precision
Agriculture
Editors: Narendra Khatri, Ajay Kumar Vyas, Celestine Iwendi, and Prasenjit
Chatterjee
Attacks on Artificial Intelligence: The New Facets of Cyber Ecospace
Editors: Kukatlapalli Pradeep Kumar, Vinay Jha Pillai, and Boppuru Rudra Prathap
ABOUT THE EDITORS
Suman Ghosal, PhD

Assistant Professor and Head of the Department of Electronics,
Pailan College of Management and Technology, West Bengal, India,
E-mail: ghosal.suman987@gmail.com
Suman Ghosal, PhD, is Assistant Professor and Head of the Department of
Electrical and Electronics Engineering at Pailan College of Management
and Technology, Kolkata, India. He is also Administrative In-Charge of
Pailan Technical Campus, Kolkata. He has more than 12 years of academic
experience in addition to industrial experience. His broad areas of interest
include renewable energy, electronics devices, and soft computing. He has
published many research papers in various national and international reputed
journals and conferences. He obtained his BTech degree from Maulana Abul
Kalam Azad University of Technology (formerly WBUT), Kolkata, and his
MTech from Bengal Engineering and Science University, Shibpur. He is
currently pursuing a PhD at the Indian Institute of Engineering Science and
Technology, Shibpu, India.
Amitava Choudhury, PhD

Assistant Professor, School of Computer Science, University of Petroleum
and Energy Studies, Dehradun, Uttarakhand, India
Amitava Choudhury, PhD, is Assistant Professor in the School of Computer
Science, University of Petroleum and Energy Studies, Dehradun, India. He
has more than ten years of experience in teaching and many years in research
work. He serves as a reviewer for IEEE Biomedical Transaction and the
journal Medical & Biological Engineering & Computing. He is member of
IEEE, UP section, and the Institute of Engineers, India. His area of research
interest is computational geometry in the field of micromechanical modelling,
pattern recognition, character recognition, and machine learning. Currently,
he is associated with the journal Cluster Computing as a guest editor. He
earned his MTech from Jadavpur University and submitted his PhD thesis at
the Indian Institute of Engineering Science and Technology, Shibpur, India.
x About the Editors
Vikram Kumar Saxena, MTech

Assistant Professor, Pailan College of Management and Technology,
West Bengal, India
Vikram Kumar Saxena is Assistant Professor at the Pailan College of Manage
ment and Technology Kolkata, India. He has been a teaching professional
for more than eight years. His research area of interest includes electrical
machines, power systems, and power electronics. He has published many
research papers in national and international journals and conferences. He
received his BTech from West Bengal University of Technology, Kolkata,
and his MTech from the National Institute of Technology Patna, India.
Arindam Biswas, PhD

Associate Professor, School of Mines and Metallurgy,
Kazi Nazrul University, West Bengal–713305, India,
E-mail: mailarindambiswas@yahoo.co.in
Arindam Biswas, PhD, is Associate Professor in the School of Mines and
Metallurgy at Kazi Nazrul University, Asansol, West Bengal, India. Dr. Biswas
has 10 years of experience in teaching research and administration. He has
published many technical papers in journals and conference proceedings as
well as six books, an edited volume, and a book chapter. Dr. Biswas was
awarded several research grants, such as from the Science and Engineering
Research Board, Govt of India, and the Centre of Biomedical Engineering,
Tokyo Medical and Dental University, in association with RIE, Shizouka
University, Japan. Presently Dr. Biswas is serving as an Associate Editor
of the journal Cluster Computing and as a guest editor of Nanoscience and
Nanotechnology and Recent Patents on Materials Science. He was formerly a
Visiting Professor at the Research Institute of Electronics, Shizouka University,
Japan. He has organized and chaired several international conferences in India
and abroad. Dr. Biswas acted as reviewer for journals and is a member of
the Institute of Engineers (India) and regular fellow of Optical Society of
India (India). His research interest is in carrier transport in low-dimensional
systems and electronic devices, non-linear optical communications, and THz
Semi-conductor sources. Dr. Biswas received an MTech degree in Radio
Physics and Electronics from the University of Calcutta, India, and a PhD
from NIT Durgapur, India. He was a postdoctoral tesearcher at Pusan National
University, South Korea, with prestigious BK21PLUS Fellowship, Republic
of Korea.
About the Editors xi
Prasenjit Chatterjee, PhD

Dean (Research and Consultancy) MCKV Institute of Engineering,
West Bengal, India
Prasenjit Chatterjee, PhD, is the Dean (Research and Consultancy) at MCKV
Institute of Engineering, India. He has published over 90 research papers
in international journals and peer-reviewed conferences. He has received
numerous awards, including Best Track Paper Award, Outstanding Reviewer
Award, Best Paper Award, Outstanding Researcher Award, and University
Gold Medal. He has been the guest editor of several special issues and has
authored and edited several books on decision-making approaches, supply
chains, and sustainability modeling. He is the lead editor of the book series
AAP Research Notes on Optimization and Decision-Making Theories and
Frontiers of Mechanical and Industrial Engineering with Apple Academic
Press. Dr. Chatterjee is one of the developers of two “multiple-criteria
decision-making methods called measurement of alternatives and ranking
according to compromise solution” (MARCOS) and “ranking of alternatives
through functional mapping of criterion sub-intervals into a single interval”
(RAFSI).
CONTENTS
Contributors......................................................................................................... xvii
Abbreviations ...................................................................................................... xxiii
Preface ............................................................................................................... xxvii
Part I: Recent Application of Deep Learning and ANN Paradigm ................... 1

1. Deep Learning Supported Evaluation of Retinal Vessel
Enhancement Techniques.............................................................................. 3
Mahua Nandy Pal, Minakshi Banerjee, and Shuvankar Roy
2. An Efficient and Optimized Deep Neural Network Architecture

for Handwritten Digit Recognition............................................................. 19
Shuvankar Roy and Mahua Nandy Pal
3. Design and Development of a Driving Assistance and

Safety System Using Deep Learning .......................................................... 35
Shekhar Verma, Tushar Bajaj, Nidhi Sindhwani, and Abhijit Kumar
4. Prediction of the Positive Cases of COVID-19 for Individual States

in India and Its Measure: A State-of-the-Art Analysis ............................. 47
Amitava Choudhury, Kaushik Ghosh, and Suman Ghosal
5. Deep Residual Neural Network Based Multi-Class Image

Classifier for Tiny Images ........................................................................... 59
Alok Negi, Prachi Chauhan, and Amitava Choudhury
6. Improved TDOA-Based Node Localization in WSN:

An ANN Approach ....................................................................................... 69
Arpan Chatterjee, Ashish Agarowala, and Parag Kumar Guha Thakurta
Part II: Role and Impact of Big Data in E-Commerce and the
Retail Sector.................................................................................................. 81
7. Impact of Big Data on E-Commerce Websites .......................................... 83
Divyang Bhartia and Mausumi Das Nath
8. A Review of the Revolutionizing Role of Big Data in Retail Industry .... 95

Mausumi Das Nath and Madhu Agarwal Agnihotri
xiv Contents
Part III: Algorithm for Load Balancing in Cloud Computing ...................... 107
9. Load-Balancing Implementation with an Algorithm of
Matrix Based on the Entire Jobs .............................................................. 109
Payel Ray, Enakshmi Nandi, Ranjan Kumar Mondal, and Debabrata Sarddar
10. Load-Balancing Algorithm with Overall Tasks....................................... 119

Enakshmi Nandi, Payel Ray, Ranjan Kumar Mondal, and Debabrata Sarddar
11. Implementation of Load Balancing of Matrix Problem with the

Row Sum Method....................................................................................... 131
Payel Ray, Enakshmi Nandi, Ranjan Kumar Mondal, and Debabrata Sarddar
Part IV: Advances in Embedded System-Based Applications ....................... 143

12. Microcontroller-Based Wearable Location Tracker System
with Immediate Support Facility.............................................................. 145
Sumanta Chatterjee, Dwaipayan Saha, Indrani Mukherjee, and Jesmin Roy
13. Design and Modifications of Eyedrop Bottle Holder Device:

A Simplest Way to Install Ocular Drug Independently.......................... 169
Sirajum Monira, Vishal Biswas, and Satyam Jha
14. Agro-Nutrition Alert (ANA)...................................................................... 187

Rajneesh Kumar, Priya Anand, Debaparna Sengupta, and Monalisa Datta
Part V: Optimization Technique Using MATLAB Platform.......................... 195

15. Selection of Optimal Mother Wavelet for Fault Analysis in
Induction Motor Using Stator Current Waveform................................. 197
Arunava Kabiraj Thakur, Palash Kumar Kundu, and Arabinda Das
16. High Efficiency High Gain SEPIC-Buck Boost Converter Based

BLDC Motor Drive for Solar PV Array Fed Water Pumping ............... 219
K. M. Ashitha and D. Thomas
17. A Classy Fuzzy MPPT Controller for Standalone PV System............... 237

Rahul Kumar Singh, Rahul Kumar Prasad, Rahul Singh, Aayush Ashish,
Debaparna Sengupta, Rishiraj Sarker, and Monalisa Datta
18. Active and Reactive Power Control in Variable Wind Turbine

Using Simultaneous Working of STATCOM and Pitch Control ........... 253
Arrik Khanna, Pushpanjali Singh Bisht, and Vikram Kumar Saxena
Part VI: Technique for Improving Information and Network Security ....... 273
19. Information Security Using Key Management ....................................... 275
Sanjib Halder, Bijoy Kumar Mandal, and Arindam Biswas
Contents xv
20. Strength-Based Novel Technique for Malicious Nodes Isolation........... 289

Prachi Chauhan and Alok Negi
21. Securing the Information Using Combined Method .............................. 305

Bijoy Kumar Mandal, Saptarshi Roychowdhury, Payel Majumder, and Anwesa Das
22. Implementation of DNA Cryptography in IoT Using Chinese

Remainder Theorem, Arithmetic Encoding, and Asymmetric
Key Cryptography ..................................................................................... 319
Ananya Satpati (Das), Soumya Paul, and Payel Majumder
23. Robust Technique to Overcome Thwarting Communication

Under Narrowband Jamming in Extremely Low Frequency of
CDMA-DSSS System on Coding- Based MATLAB Platform ............... 333
Tanajit Manna, Rahit Basak, Mainak Das, and Alok Kole
Part VII: Miscellaneous Topics in Computing ................................................ 347

24. The Utility of Regional Language Facility in Accounting
Software for Retail Market of West Bengal............................................. 349
Madhu Agarwal Agnihotri and Arkajyoti Pandit
25. Comparative Analysis of Storey Drift of a G+9 RC-Structured

Building Due to Seismic Loads in Various Seismic Regions .................. 363
Budhaditya Dutta, Aditya Narayan Chakraborty, Ayushi Shah, Debarka Brahma,
Amit Deb, and Tushar Kanti De
26. A Proposal for Disinfection of Contaminated Subject Area

Through Light ............................................................................................ 383
Anubrata Mondal and Kamalika Ghosh
Index .................................................................................................................... 397

CONTRIBUTORS
Ashish Agarowala
CSE Department, NIT Durgapur, West Bengal, India
Madhu Agarwal Agnihotri

Assistant Professor, Department of Commerce, St. Xavier’s College (Autonomous), Kolkata, West Bengal,
India, E-mails: madhu.cal@gmail.com; madhu.cal@sxccal.edu
Priya Anand
Department of Information Science and Engineering, AMC Engineering College, Bangalore–560083,
Karnataka, India
Aayush Ashish
Techno International New Town, Kolkata–700156, West Bengal, India
K. M. Ashitha
Department of Electrical and Electronics Engineering, Amal Jyothi College of Engineering, Kanjirappally,
Kerala–686518, India, E-mail: ashithakm@ee.ajc.i
Tushar Bajaj
Department of ECE, Amity School of Engineering and Technology, New Delhi GGSIPU, Delhi, India
Minakshi Banerjee
Computer Science and Engineering Department, RCC Institute of Information Technology,
Kolkata–700015, West Bengal, India
Rahit Basak
Computer Science and Engineering Department, Pailan College of Management and Technology,
Divyang Bhartia
Student, St. Xavier’s College (Autonomous), Kolkata, 30 Mother Teresa Sarani, Kolkata–700016,
West Bengal, India
Pushpanjali Singh Bisht

Department of Electrical Engineering, DPGITM Gurugram, Haryana–122001, India
Arindam Biswas
Kazi Nazrul University, Asansol, West Bengal, India
Vishal Biswas
Department of Bachelor in Optometry, Pailan College of Management and Technology, Joka,
Debarka Brahma
Undergraduate Student, Department of Civil Engineering, Amity University Kolkata, Kolkata,
West Bengal–700135, India
Aditya Narayan Chakraborty

xviii Contributors
Arpan Chatterjee
Sumanta Chatterjee
Department of Computer Science and Engineering, JIS College of Engineering, Kalyani, West Bengal,
India, E-mail: Sumanta.chatterjee@jiscollege.ac.in
Prachi Chauhan
G.B. Pant University of Agriculture and Technology, Pantnagar, Uttarakhand, India,
E-mail:cprachi664@gmail.com
Amitava Choudhury
School of Computer Science, University of Petroleum and Energy Studies, Dehradun, Uttarakhand,
India, E-mail: a.choudhury2013@gmail.com
Anwesa Das
NSHM Knowledge Campus, Durgapur, West Bengal, India
Arabinda Das
Department of Electrical Engineering, Jadavpur University, Kolkata–700032, West Bengal, India
Mainak Das
Computer Science and Engineering Department, Pailan College of Management and Technology,
Monalisa Datta
Department of Electrical Engineering, Techno International New Town, Kolkata–700156, West Bengal,
India
Tushar Kanti De
Associate Professor and Head of Department, Department of Civil Engineering,
Budge Budge Institute of Technology, Budge Budge, West Bengal–700137, India
Amit Deb
Assistant Professor, Department of Civil Engineering, Amity University Kolkata, Kolkata,
Budhaditya Dutta
West Bengal–700135, India, E-mail: bdneil3103@gmail
Suman Ghosal
Department of Electrical and Electronics Engineering, Pailan College of Management and Technology,
Kolkata, West Bengal, India
Kamalika Ghosh
Director, School of Illumination Science, Engineering and Design, Jadavpur University, Kolkata,
West Bengal, India, E-mail: kamalikaghosh4@gmail.com
Kaushik Ghosh
School of Computer Science, University of Petroleum and Energy Studies, Dehradun, Uttarakhand, India
Sanjib Halder
The Bhawanipur Education Society College, Kolkata, West Bengal, India
Satyam Jha
Contributors xix
Arrik Khanna
Department of Electrical Engineering, Chitkara University, Punjab–140401, India,
E-mail: arrik1433@gmail.com
Alok Kole
Electrical Engineering Department, RCC Institute of Information Technology, Kolkata–700015,
West Bengal, India
Abhijit Kumar
School of Computer Science, University of Petroleum and Energy Studies, Dehradun, Uttarakhand,
India, E-mail: abhijitkaran@hotmail.com
Rajneesh Kumar
India, E-mail: rajneeshkumar4044@gmail.com
Palash Kumar Kundu

Department of Electrical Engineering, Jadavpur University, Kolkata–700032, West Bengal, India
Payel Majumder
Department of Computer Science and Engineering, NSHM Knowledge Campus, Durgapur,
Arrah Shibtala Via Muchipara, Durgapur–713212, West Bengal, India
Bijoy Kumar Mandal

NSHM Knowledge Campus, Durgapur, West Bengal, India, E-mail: writetobijoy@gmail.com
Tanajit Manna
Electronics and Communication Engineering Department, Pailan College of Management and
Technology, Kolkata–700104, West Bengal, India, E-mail: tanajitmanna@gmail.com
Anubrata Mondal
Research Scholar, School of Illumination Science Engineering and Design,
(Jadavpur University, Kolkata) and Assistant Professor, Department of Electrical Engineering,
Global Institute of Management and Technology, West Bengal, India
Ranjan Kumar Mondal

Department of Computer Science and Engineering, University of Kalyani, West Bengal, India,
E-mail: ranjangcett@gmail.com
Sirajum Monira
Kolkata–700104, West Bengal, India, E-mail: srisikhsa@gmail.com
Indrani Mukherjee
India
Enakshmi Nandi
E-mail: nandienakshmi@gmail.com
Mausumi Das Nath

Assistant Professor, Department of Commerce, St. Xavier’s College (Autonomous), Kolkata,
30 Mother Teresa Sarani, Kolkata – 700016, West Bengal, India, E-mail: m.dasnath@sxccal.edu
Alok Negi
National Institute of Technology, Uttarakhand, India
xx Contributors
Mahua Nandy Pal

Computer Science and Engineering Department, MCKV Institute of Engineering, Howrah–711204,
West Bengal, India, E-mail: mahua.nandy@gmail.com
Arkajyoti Pandit
Assistant Professor, Department of Commerce, St. Xavier’s College (Autonomous), Kolkata,
West Bengal, India
Soumya Paul
Department of Computer Science and Engineering, NSHM Knowledge Campus, Durgapur,
Arrah Shibtala Via Muchipara, Durgapur–713212, West Bengal, India
Rahul Kumar Prasad

Payel Ray
E-mail: payelray009@gmail.com
Jesmin Roy
India
Shuvankar Roy
Computer Science and Engineering Department, MCKV Institute of Engineering, Howrah–711204,
West Bengal, India, E-mail: shuvankarroy2@gmail.com
Saptarshi Roychowdhury
NSHM Knowledge Campus, Durgapur, West Bengal, India
Dwaipayan Saha
India
Debabrata Sarddar
E-mail: dsarddar1@gmail.com
Rishiraj Sarker
Jadavpur University, Kolkata–700032, West Bengal, India, E-mail: sarker.rishiraj88@gmail.com
Ananya Satpati (Das)

Department of Computer Science and Engineering, NSHM Knowledge Campus, Durgapur, Arrah
Shibtala Via Muchipara, Durgapur–713212, West Bengal, India, E-mail: ananya.das@nshm.com
Vikram Kumar Saxena

Department of Electrical Electronics and Engineering, Pailan College of Management and Technology,
Debaparna Sengupta
India
Ayushi Shah
Nidhi Sindhwani
Contributors xxi
Rahul Kumar Singh

Rahul Singh
Arunava Kabiraj Thakur

Department of Electrical Engineering, Techno Main Salt Lake, Sector V, Bidhannagar, Kolkata–700091,
West Bengal, India, E-mail: arunava.kabiraj007@gmail.com
Parag Kumar Guha Thakurta

D. Thomas
Amal Jyothi College of Engineering, Kanjirappally, Kottayam, Kerala–686518, India
Shekhar Verma
ABBREVIATIONS
2D two dimensional
ACC adaptive cruise control
AGC adaptive gamma correction
AHE adaptive histogram equalization
ANA agro nutrition alert
AWGN additive white gaussian noise
BER bit error rate
BS base station
BT body temperature
CDF cumulative distribution function
CDMA code division multiple access
CLAHE contrast limited adaptive histogram equalization
CNN/ConvNet convolutional neural network
COG center of gravity
Coif coiflets
CQA classification quality assessment
CSA control based pitch actuator system
CTC connectionist temporal classification
db Daubechies
DES data encryption standard
DGR directional greedy routing
DL dead load
DNA deoxyribonucleic acid
DNN deep neural network
DSSS direct sequence spread spectrum
DWT discrete wavelet transform
EBGR edge node based greedy routing
ESU energy storage unit
EVCSs electric vehicle charging stations
EVs electric vehicles
FFT fast Fourier transform
FLCs fuzzy logic controllers
FN false negative
FP false positive
xxiv Abbreviations
GPS global positioning system

GSM global system for mobile communications
GWEC Global Wind Energy Council
GyTAR greedy traffic-aware routing
HOG histograms of oriented gradients
HR heart rate
HRV heart rate variability
HUD heads up display
I/O input/output
ICP ideal customer profile
IDWT inverse discrete wavelet transform
IoT internet of things
IQA image quality assessment
IR infrared
IRB institutional review board
KNN k-nearest neighbor
LBMM load balance Min-Min
LED light emitting diode
LiDAR light detection and ranging
LL live load
LSTM long short term memory
LVRT low voltage ride through
MCT minimum completion time
MM min-min
MNIST modified national institute of standards and technology
MOM mean of maximu m
MPPT maximum power point tracking
MSE mean squared error
NB negative big
NS negative small
OLB opportunistic load balancing
P&O perturb and observe
PB positive big
PS positive small
PSD power spectral density
PSNR peak signal to noise ratio
ReLU rectified linear unit
RMSE root of mean square error
RNA ribonucleic acid
Abbreviations xxv
RNN recurrent neural network

ROI region of interest
RSS received signal strength
RSU road side unit
SD standard deviation
SGD stochastic gradient descent
SGDM stochastic momentum gradient descent
SISED School of Illumination Science and Engineering
Sn sensitivity
STATCOM static compensator
STC standard test conditions
STFT short term Fourier transform
SVM support vector machine
Sym Symlets
TDOA time difference of arrival
TN true negative
TP true positive
TSO transmission system handling operators
UV ultraviolet
V voltage
V2V vehicle to vehicle
VSI voltage switching inverters
VUV vacuum UV
WD weighting distribution
WHO World Health Organization
WSN wireless sensor network
ZE zero
PREFACE
This book aims to provide a sharing platform for researchers, academicians,

and industry to throw in their expertise, acquaintance, and understanding in
a wider domain of progressive and emerging technologies, including arti
ficial intelligence, communication, cyber security, data analytics, Internet
of Things (IoT), machine learning, power system, VLSI, embedded system,
and much more. It aims to be a transdisciplinary and cross-cutting resource
that present recent trends and advances and challenges.
We are delighted to deliver this book with the support and contributions
from global academicians and researchers. This book is a manifestation of
various interesting and important aspects of data science and computing
technologies and their applications and methodologies in a wider range
of applications including deep learning, DNA cryptography, classy fuzzy
MPPT controller, driving assistance, and safety systems.
This book consists of 26 chapters, covering discussion on innovative
algorithms and their applications for solving cutting-edge computational and
data science problems with an interdisciplinary research perspective. All 26
chapters are divided in following seven parts containing interdisciplinary
applications and methodologies of data science and computing technologies:
 Part I: Recent Application of Deep Learning and ANN Paradigm
 Part II: Role and Impact of Big Data in E-Commerce and Retail Sector
 Part III: Algorithm for Load Balancing in Cloud Computing
 Part IV: Advances in Embedded System Based Application
 Part V: Optimization Technique Using MATLAB Platform
 Part VI: Technique for Improving Information and Network Security
 Part VII: Miscellaneous Topics in Computing
In summary, in this 21st century, there is a genuine need for technology
upgradation, and this book provides interdisciplinary research and a techno
logical overview that the editors hope will lead to open a new dimension of
understanding, which will be useful for uncovering new in this growing area.
The editors of this book are thankful to all authors, reviewers, and Apple
Academic Press for their valuable input and contributions to make this book
possible.
—Editors
PART I
Recent Application of Deep Learning
and ANN Paradigm
CHAPTER 1
DEEP LEARNING SUPPORTED

EVALUATION OF RETINAL VESSEL
ENHANCEMENT TECHNIQUES
MAHUA NANDY PAL,1 MINAKSHI BANERJEE,2 and
SHUVANKAR ROY1
1
Computer Science and Engineering Department,
MCKV Institute of Engineering, Howrah–711204, West Bengal, India,
E-mail: mahua.nandy@gmail.com (M. N. Pal)
2
Computer Science and Engineering Department, RCC Institute of
Information Technology, Kolkata–700015, West Bengal, India
ABSTRACT
Hypertension, diabetes, glaucoma, etc., are different diseases, the symptoms

of which prevails in retinal fundus images at preliminary stage. Retinal
vessel segmentation is a significant prior step for different automated disease
analysis system of retinal fundus images. Further, image enhancement has a
fundamental role in retinal vessel segmentation. The better is the enhance
ment the better segmentation result is expected. In this work, three enhance
ment techniques are considered which are very frequently used before retinal
vessel segmentation in contemporary literatures. The competency of each of
the enhancement techniques is evaluated using two different categories of
image quality assessment (IQA) metrics, peak signal to noise ratio (PSNR)
and absolute mean brightness error (AMBE). In this way, morphological
operation performs the best when compared with other two enhancement
techniques. This observation has been further evaluated from a reverse view
Advances in Data Science and Computing Technology: Methodology and Applications. Suman Ghosal,
Amitava Choudhury, Vikram Kumar Saxena, Arindam Biswas, & Prasenjit Chatterjee (Eds.)
© 2023 Apple Academic Press, Inc. Co-published with CRC Press (Taylor & Francis)
4 Advances in Data Science and Computing Technology
point of effective deep convolutional neural network (CNN) vessel classifi

cation result on three differently enhanced image dataset which supports the
previous observation in terms of classification efficiency metrics as well.
Better vessel classification metrics are obtained from the same classification
model when input datasets are better enhanced to represent vessel.
1.1 INTRODUCTION
Retinal fundus image is characterized both by anatomical and pathological

artifacts present in its surface. The key anatomical artifacts are blood vessels,
optic disc, and fovea. Pathological artifacts are usually symptoms of different
optical and cardio-vascular diseases. These are manifested at an early stage
in retinal image surface. Automated retinal image analysis system helps
medical professionals in fast screening of huge real life dataset. Challenges
present in automated retinal diagnostic system are presence of noise, low
contrast, uneven illumination, minute sizes of medically significant local
artifacts such as micro aneurysms, hemorrhages, etc. In process of automated
retinal analysis system development, retinal vessel segmentation plays a
significant role, Moreover, retinal image enhancement is the pre requisite of
retinal vessel segmentation.
1.2 LITERATURE SURVEY
Adaptive histogram equalization (AHE) is an image enhancement

technique. But AHE reports two major problems namely slow speed
and over enhancement of noise. These problems are addressed by Pizer
et al. (1987). Zuiderveld in 1994 presented contrast limited adaptive
histogram equalization (CLAHE), to acquire required results overcoming
the disadvantages Nandy and Banerjee (2017) made use of CLAHE for
enhancement of images before vessel segmentation. Huang, Cheng, and
Chiu (2013) dealt with AGC based image enhancement. They presented
image transformation through gamma correction and probability distribution
of luminance pixels. Rahman et al. (2016) discussed application of AGC
for different types of image enhancement. An adaptive gamma correction
(AGC) based on image contrast is adopted in Nandy and Banerjee (2019).
Application of morphological operation has been utilized in medical image
enhancement in 2015 (Hassanpour et al., 2015). Hassan et al. (2015) used
different morphological operators and clustering for vessel segmentation. An
Deep Learning Supported Evaluation of Retinal Vessel 5
entropy based image quality metric has been worked out in 2012 (Nandy
and Banerjee, 2012). Sheba and Gladston (2016) evaluated different image
enhancement techniques in the domain of mammogram images.
Almotiri et al. (2018) presented a survey on different retinal vessel
segmentation algorithms. Roychowdhury et al. (2015) iteratively thresholded
the residual retinal image to obtain higher vessel segmentation accuracy. Fu
et al. (2016); Liskowski et al. (2016); and Mo and Zhang (2017) exploited
deep learning architecture to identify vessels from retinal fundus images. Fu
et al. (2016) used convolutional neural network (CNN) which follows vessel
probability map. In this chapter, fully connected conditional random fields
are combined with probability map to segment vessels. Liskowski et al.
(2016) applied several preprocessing steps and training was conducted on a
large number of samples. Generally large volume of training data is required
for effective learning of deep architecture. These processes are extremely
computationally exhaustive and the training is also very time consuming
with requirements of expensive resources.
Ronneberger et al. (2015) proposed U-net CNN model which is efficient
in segmenting bio medical images.
The contribution is to find the effective enhancement technique quantita
tively following two diverse viewpoints. We have experimented with some
of the enhancement techniques, frequently used for retinal vessel segmenta
tion such as, CLAHE, AGC and morphological enhancement techniques. On
the basis of IQA metric values, we have established with the notion that
morphological enhancements are the best among them in an application
oriented way. To arrive at a conclusive result from a different dimension, we
experimented with the efficiency of deep vessel classification on three differ
ently enhanced image datasets also. In this phase, the better the classification
evaluation metric values the more informative the input images are. The
results obtained from deep classification experiments are decently aligned
with the observations obtained from image quality assessment (IQA) too.
The work representation is structured as follows. Section 1.3 presents
enhancement methods normally used before vessel segmentation of retinal
fundus. IQA evaluation metrics are there in Section 1.4 and deep classifica
tion based evaluation metrics. Section 1.5 presents database description.
Section 1.6 explores deep neural network (DNN) architecture, training
details and implementation requirements used to evaluate our observation.
Section 1.7 describes experimental results of IQA based evaluation and deep
CNN classification based evaluation respectively. Section 1.8 demonstrates
the experimental results found in Section 1.7. Section 1.9 is the conclusion
of the work.
1.3 IMAGE ENHANCEMENT TECHNIQUES
1.3.1 CONTRAST LIMITED ADAPTIVE HISTOGRAM-BASED

ENHANCEMENT
General image transformation function is:

=s T ( r ) ; 0 ≤ r ≤ lmax (1)
The lmax is maximum intensity value.
Pixel intensity levels of an image may be thought of as random variables.
Random variables are fundamentally described by probability density func
tion. Thus, the transformation function is:
r
= (r )
s T= ∫0
pr ( w ) dw (2)
where; w is a dummy integration variable. Eqn. (2) represents the cumulative
distribution function (CDF) (digital image processing, Gonzalez). Following
discrete representation, the probabilities of occurrence of intensity level rk is:
nk
Pr ( rk ) = (3)
MN
where; nk is the number of pixels having intensity level rk. So, the discrete
variation of the function is:
( lmax )
= ( rk )
sk T= (4)
MN ∑ j =0n j
k
where; k = 0, 1, …, lmax
Eqn. (4) represents histogram equalized image. Pizer et al. (1987) elabo
rated AHE application. Zuiderveld (1994) modified it to CLAHE to obtain
better enhancement result. CLAHE follows local region based approach. All
the local regions contrast enhancement takes place independently and are
combined with bilinear interpolation. Thus, boundary effects between the
region can be removed.
1.3.2 ADAPTIVE GAMMA CORRECTION (AGC)-BASED

ENHANCEMENT
AGC with weighting distribution (WD) (Huang, Cheng, and Chiu, 2013) is
depicted as follows. Gamma correction formula is:
γ
 l 
T ( lmax ) = ( lmax )   (5)
 lmax 
where; lmax is the maximum intensity. The gamma value may be greater than
1 or may be less than 1 (Huang et al., 2013). AGC with weight distribution
is as follows (Huang et al., 2013):
1−cdf ( l )
T ( l ) =lmax ( l / lmax ) (6)
The underlined concept is that the process tries to increase lower intensity
level and at the same time tries to resist decrease of higher intensity level.
The WD function has the following formula:
pdf ( l ) − pdf min
pdf w ( l ) = pdf max ( )a (7)
pdf max − pdf min
where; a is the adjusted parameter. Thus, cdf becomes:
pdf w ( l )
cdf w ( l ) = ∑ lmax
l
(8)
=0
∑ pdf w
where; Ʃpdfw is defined as:
∑ pdf = ∑ pdf w ( l )
lmax
w l
(9)
At the end, the gamma (γ) value is obtained in the following way:
γ = 1− cdf w ( l ) (10)
1.3.3 MORPHOLOGICAL OPERATION-BASED ENHANCEMENT
Morphological operations make use of structuring elements while enhancing

retinal artifacts. Disk shaped masks are generally used among researchers in
biomedical image data enhancement. Tophat and bottom-hat transformations
are represented (Digital Image Processing, Gonzalez) in Eqns. (11) and (12),
respectively:
Tophat ( I=
) ( I − ( I ( open ) b ) ) (11)
Bottomhat
= ( I ) ( ( I ( close ) b ) − I ) (12)
where; b is structuring element and morphological operation (I(open)b)is

erosion of I by b, followed by dilation with b and (I(close)b is dilation of
I by b, followed by erosion with b. According to Hassanpour et al. (2015),
subsequent applications of top-hat and bottom-hat transformation lead to

contrast enhancement between light and dark arenas.
I + Tophat ( I ) − Bottomhat ( I )
En = (13)
When Eqn. (13) is tried with different structural elements of suitable size
enhancement result varies. The output enhanced retinal image is (En).
1.4 EVALUATION METRICS
We used two types of evaluation metrics for the assessment of retinal image
enhancement techniques. They are IQA metrics and classification quality
assessment (CQA) metrics.
1.4.1 IQA METRICS
IQA metrics are categorized in two classes: statistical error based metrics and
visual information based metrics. Ideally, a human observer only can provide
most consistent IQA metric to assess an image enhancement performance. It
is qualitative evaluation. Visual information may not be properly represented
by statistical error based metrics always instead human observer perception
can be quite aptly captured by visual information based metrics. PSNR and
AMBE are the mostly used statistical error based IQA metric and visual
information based IQA metric, respectively.
1.4.1.1 PEAK SIGNAL TO NOISE RATIO (PSNR)
PSNR is defined as:

( lmax )
2
PSNR = 10log (14)

MSE
MSE is mean squared error (MSE) between the original image(I) and the
enhanced image(En) and is defined as:
1
∑ ( En ( m,n ) − I ( m,n ) )
2
=MSE (15)
MN m,n
Higher PSNR value specifies more information content.

1.4.1.2 ABSOLUTE MEAN BRIGHTNESS ERROR (AMBE)
AMBE is the absolute difference between enhanced image expectation

(E(En)) and original image expectation (E(I)).
AMBE
= ( E ( En ) − E ( I ) ) (16)
Lower AMBE value specifies better image enhancement.
1.4.2 DEEP CLASSIFICATION METRICS
The significances of true positive (TP), false positive (FP), true negative
(TN), and false negative (FN) with respect to vessel classification are as
follows, TP is the proportion of actual vessel pixels in the whole popula
tion whereas, TN is the proportion of actual non-vessel pixels in the whole
population. Similarly, FP is the proportion of falsely identified vessel pixels
in the whole population whereas FN is the proportion of falsely recognized
non-vessel pixels in the whole population. Accuracy and sensitivity are two
important evaluation metrics in case of medical image processing.
1.4.2.1 SENSITIVITY
Sensitivity (Sn) is TP rate. The closer the value of sensitivity to 1, the better
is the identification power of vessel pixel of the method.
=Sn TP / (TP + FN ) (17)
1.4.2.2 ACCURACY
Accuracy (Acc) provides proportion of both TP and true negative.

Acc = (TP + TN ) / (TP + TN + FP + FN ) (18)
1.5 DATASET DESCRIPTION
DRIVE (Digital retinal image for vessel extraction, Image Sciences Insti
tute) retinal image vessel segmentation dataset is used for method evalu
ation as it is the most frequently used dataset among researchers. Canon
CR5 non-mydriatic 3CCD camera with 45° field of view has been used
to capture retinal images. Each image was preserved with 8 bits per color
plane and at 565 x 584 pixels’ resolution. In this dataset ground truths are
also available.
1.5.1 PATHOLOGICAL SAMPLES
Around 40 photographs have been randomly considered, 33 are healthy

and 7 show symptoms of mild early diabetic retinopathy. Among all those
7 images 3 belongs to training set whereas 4 belongs to test set. A brief
description of the abnormalities is as available online (https://drive.grand
challenge.org/):
• 25 Training: Pigment epithelium changes, probably buttery macu
lopathy with pigmented scar in fovea, or choroidopathy, no diabetic
retinopathy or other vascular abnormalities.
• 26 Training: Background diabetic retinopathy, pigmentary epithelial
atrophy, atrophy around optic disk.
• 32 Training: Background diabetic retinopathy.
• 03 Test: Background diabetic retinopathy.
• 08 Test: Pigment epithelium changes, pigmented scar in fovea, or
choroidopathy, no diabetic retinopathy or other vascular abnormalities.
1.6 DEEP NEURAL NETWORK (DNN)
A deep CNN model has been implemented and then tested with differently
pre-processed image datasets. CNN architecture has been formed using U-net
type connections. U-net is a CNN, efficient in the field of medical image
segmentation. The characteristic of U-net is that it utilizes the same feature
maps for contraction of the image to a feature vector and for expansion of
the vector to a reconstructed image. Thus, structural integrity of image data
is retained. U-net model is trained on DRIVE trainset for vessel identifica
tion. Application of U-net model on appropriately enhanced image dataset
leads to pixel level classification of vessel and non-vessel. Classification
efficiency measurement of the same model with differently enhanced image
dataset indicates better image enhancement method.
1.6.1 ARCHITECTURE
U-net architecture Ronneberger et al. (2015) is used here for CNN based
input IQA. Cross-entropy loss function and stochastic gradient descent
(SGD) optimization algorithm is used in implementation. Rectified linear
unit (ReLU) is used as activation function. A drop out of 0.2 has been
employed between two consecutive convolutional layers. The summary of
the train model is represented in Figure 1.1.
FIGURE 1.1 Summary of the deep CNN model.

1.6.2 TRAINING
DRIVE images undergo three types of image enhancement techniques

to obtain three different datasets as discussed in Section 1.3. Training is
performed on image patches of differently preprocessed image datasets.
Data augmentation has not been performed. During deep classification based
evaluation, 48 x 48 random overlapping patches are generated from image
dataset for proper training with less number of image samples. 90% of patch
dataset has been used for training whereas 10% of patch dataset has been
used for validation. Mini batch size is 32 patches.
1.6.3 IMPLEMENTATION REQUIREMENTS
All the experiments have been implemented in Python 3.6.8. High level APIs
of Google’s deep learning library TensorFlow have been used to implement
and evaluate the work. Implementations have been executed in Google
online Colab cloud environment where Tesla K80 GPU, 12 GB VRAM, 4*
Intel(R) Xeon(R) CPU @ 2.20 GHz, 13.51 GB RAM, 358.27 GB HDD are
available as deep network execution resources.
1.7 EXPERIMENTAL RESULTS
1.7.1 IMAGE QUALITY ASSESSMENT (IQA) METRIC BASED

EVALUATION
During enhancement, green channel of retinal image is only considered

(Nandy and Banerjee, 2012) as the intensity in green channel facilitates
retinal structures to be observable. Retinal images are almost saturated in
the red channel and have very low contrast in blue channel. Green channel
images are contrast enhanced by applying CLAHE. Another effective image
enhancement tool is Gamma correction. Green channel images undergo
adaptive linear gamma correction where all the pixels are linearly weighted
by some predefined factor (Huang et al., 2013). One more exhaustively used
retinal image enhancement tool is application of morphological operators
(Hassanpour et al., 2015). All four images and their histograms are given in
Figure 1.2. Average PSNR and AMBE for DRIVE images are shown both
graphically and in tabular form in Figure 1.3 and Table 1.1, respectively.
FIGURE 1.2 (a) Retinal green channel and histogram; (b) CLAHE and stretched histogram;
(c) morphology and stretched histogram; (d) AGC and stretched histogram.
FIGURE 1.3 Average PSNR and AMBE graph for DRIVE images.
TABLE 1.1 DRIVE Test Set Images – Average IQA Metrics Value
Enhancement Avg Image Quality Assessment Metrics Value
Techniques PSNR AMBE
Morphology 24.8719 5.7424
CLAHE 14.3472 35.6176
AGC 12.6987 47.6397
1.7.2 DEEP CLASSIFICATION METRIC BASED EVALUATION
48 x 48 random and overlapping image patches are generated from image

dataset. U-net type deep CNN architecture has been fed with patches for
training. A total of 1,71,000 training patches and 19,000 validation patches

were generated from DRIVE train set images. The model was trained for 10
epochs only, as the trend of classification result can be observed from less
number of epochs also. Accuracy and sensitivity are two important evalua
tion metrics in the field of vessel classification. Average training accuracy,
test accuracy and sensitivity obtained from differently enhanced image
datasets are in Table 1.2. Graphs for average training and test accuracies
and receiver operator characteristic graph of corresponding classification are
also presented from Figure 1.4 to Figure 1.6.
TABLE 1.2 DRIVE Test Set Images – Average Deep Classification Metrics
Enhancement Avg Classification Quality Assessment Metrics Value
Techniques Train_Acc Test_Acc Sensitivity
Morphology 0.9576 0.9524 0.7573
CLAHE 0.9568 0.9525 0.7564
AGC 0.9510 0.9484 0.7349
FIGURE 1.4 ROC and accuracy curves of morphologically enhanced images.
FIGURE 1.5 ROC and accuracy curves of CLAHE images.

FIGURE 1.6 ROC and accuracy curves of AGC enhanced images.
1.7.2 IMAGE QUALITY ASSESSMENT (IQA) OF PATHOLOGY

SAMPLES
Four images among 20 drive test images have mild early DR signs as patho
logical symptoms. Table 1.3 shows individual performance measurements
obtained by applying deep CNN classification on those pathology samples
separately. This is a reverse view point to assess the applicability of different
enhancement techniques on abnormal images also. In most of the cases
morphologically enhanced abnormal images are characterized with higher
metric values.
TABLE 1.3 Deep CNN Classification Metrics on DRIVE Test Dataset Abnormal Images
Symptoms Enhancement Deep Classification Assessment Metrics
Techniques Accuracy Sensitivity
Hard exudate Morphology 0.9420 0.6921
CLAHE 0.9420 0.6928
AGC 0.9420 0.7670
Hard exudate Morphology 0.9464 0.6580
CLAHE 0.9458 0.6489
AGC 0.9365 0.633
Hemorrhage Morphology 0.9574 0.8164
CLAHE 0.9571 0.8106
AGC 0.9535 0.7987
hard exudate Morphology 0.9493 0.6864
CLAHE 0.9515 0.7153
AGC 0.9487 0.7082
1.8 DISCUSSION
Statistical error based and visual perception based IQA metrics are
considered in this work for preliminary image enhancement evaluation.
Here, PSNR represents statistical error based metrics. But it is much more
important to verify evaluation with human visual perception based metric
such as AMBE. Maximum PSNR value specifies high information content
of the morphologically enhanced image. Again, minimum AMBE value of
morphologically enhanced image also identifies it as better enhanced image.
Table 1.1 data represents the fact properly. Average values are computed on
DRIVE test dataset images.
This observation is further estimated from the view point of deep CNN
classification as vessel enhancement is a very important prior sub part of
retinal vessel segmentation. So, better enhanced images tend to produce better
classification efficiency metrics. Maximum value of average segmentation
accuracy indicates the superior quality of morphologically enhanced image data
with application to retinal fundus images. In the field of vessel segmentation
use of sensitivity as performance measurement index is highly significant.
Maximum average sensitivity has also been obtained from morphologically
enhanced test images. Table 1.2 represents average values of train accuracy,
test accuracy and sensitivity on DRIVE images. Different ailment symptoms
are evident in retina at a very early phase. Table 1.3 provides individual test
accuracy and sensitivity of pathology samples of DRIVE test set separately.
In individual evaluation of pathology samples also, most of the classification
efficiency values are in favor of morphological operation.
Though morphological enhancement performed very well in both IQA
based evaluation as well as deep classification based evaluation, performing
image enhancement by combining enhancement methods in some specific
way may produce better result as well.
1.9 CONCLUSION
Three single image enhancement techniques are experimented in this

work for their efficiency in image enhancement with application to retinal
vessel segmentation. These single enhancement techniques are applica
tions of CLAHE, AGC, and morphological enhancements. From the view
point of IQA metric values it is decided quantitatively that morphological
enhancement performs well for this purpose. This claim has been further
established by deep CNN vessel classification based evaluation also as better
Another random document with
no related content on Scribd:
separation for years; then this sudden need to meet him again in
order to side-track some ‘embarrassments’: obviously she had
married him, and needed a divorce if she was to marry again. I wish
most problems were as simple.”
“And, of course, if she married Staveley in 1915, as she seems to
have done, he committed bigamy in marrying Mrs. Fleetwood?”
“Which means that Mrs. Fleetwood is Mrs. Fleetwood, and that
she's legally married now. She won't be sorry to hear it. That's why I
sent Mme. Laurent-Desrousseaux up there now. First-hand evidence
is better than documents; and, of course, the documents will be
forthcoming if they're required in future. Evidently those three
scoundrels didn't know this latest twist in the affair, or they wouldn't
have tried the trick they did last night. They'd have done worse,
probably, when they got hold of her. If she'd been dead and out of
the way, there would have been no one except old Miss
Fordingbridge to contest that impostor's claim—and she was so
besotted with him that she'd never have dreamed of doing so.”
He paused for a moment or two, as though considering the case;
but when he spoke again it was on a different point.
“You sometimes jeer at me for playing the mystery-man and
refusing to tell you what I infer from the facts that turn up. It's
sometimes irritating, I admit; and now and again I suppose it makes
me look as if I were playing the superior fellow. But it's really nothing
of the sort. In affairs of this kind, one never can tell what the next
turn of the wheel may be; and one might quite well blurt out
something which would give the cue to the very people you want to
keep in the dark.”
“You do irritate me often enough, Clinton,” Wendover admitted. “I
can't see why you shouldn't put your cards on the table. A fact's a
fact, after all.”
“I'll give you just one example,” said Sir Clinton seriously.
“Suppose I had blurted out the fact which I'd inferred about Mme.
Laurent-Desrousseaux's marriage. It was implicit in the story she told
us; but luckily no one spotted the key except myself. Now, just think
what would have happened to-night if that had been common
property. These scoundrels would have known that Mrs. Fleetwood
was legally married to young Fleetwood, since the ceremony with
Staveley was illegal. Therefore, instead of trying the business of the
forced marriage, they'd simply have pitched her over the cliff at the
Blowhole. She'd have been dead by this time; for their only interest
in keeping her alive was to force this marriage with the claimant and
side-track difficulties in that way. Suppose I'd blurted out my
inference, and sent that girl to her death by my carelessness, how
should I be feeling at this moment? None too comfortable, so far as I
can see.”
Wendover had to admit that the secrecy policy had justified itself.
“It would have been a dreadful business,” he confessed.
Inspector Armadale's figure appeared from one of the corridors,
and, catching sight of Sir Clinton, he came over to where they were
standing. His face showed that he had good news to tell.
“I've got practically the whole business out of them, sir. Billingford
gave everything away that he knew about; and the other chap's
nerve was completely gone, so that he couldn't resist questioning.
It's as clear a case as one could wish for.”
He paused, as though puzzled by something, and then added:
“It beats me how you tumbled to the fact that Cargill was one of
the gang, though.”
Sir Clinton ignored the underlying inquiry.
“Was he the brains of the show?” he asked. “I've only a suspicion
to go on there.”
“Yes, he did the planning for them.”
“And the gentleman with no face collaborated with Aird in the
actual murders? That's a guess, I may say, so far as the Staveley
affair's concerned, though I'm fairly sure of my ground in the other
cases.”
“You're right in that case too, sir. Aird and the impostor fellow
were the actual murderers. Aird'll hang for certain.”
“It'll make a very nice case for you, inspector; and I'm sure you'll
work it up well for the Public Prosecutor. I can seen a laurel wreath
somewhere in the background.”
“But it's you who did most of it, sir. Nobody understands that
better than I do,” the inspector objected, evidently afraid lest Sir
Clinton thought him capable of accepting the credit without protest.
“I came into the thing on a strict understanding that I was to be a
pure spectator, you remember. I'm afraid that at times I got a shade
too zealous, perhaps; but it's your case and not mine. If we'd made a
mess of it between us, you'd have had to stand the racket; so
obviously a success goes down to your account. The subject's
closed.”
Wendover, seeing the inspector's difficulty in framing a suitable
reply to this, intervened to change the subject.
“I see the main outlines of the affair easily enough, Clinton,” he
said, “but I'd like to hear just how you worked it out as you went
along. Any objections to telling me? It'll go no farther, of course.”
The chief constable's face betrayed a tinge of boredom.
“You've lived with this case for the best part of a week. Haven't
you had enough of it by this time?”
Wendover persisted in his demand; but Sir Clinton, instead of
complying, glanced at his watch.
“There's one detective story I'm very fond of, squire: The Hunting
of the Snark. I rank it high in the scale, especially on account of the
number of apt quotations one can make from it. Here's one:
The method employed I would gladly explain

While I have it so clear in my head,
If I had but the time and you had but the brain—
But much yet remains to be said.
It's far too late to start a long story to-night. I'm dead sleepy. If
you remind me about it to-morrow, I'll do my best; but I will not sit up
all night even to please you.”
The inspector seemed as much disappointed as Wendover at his
superior's decision.
“I'd like to hear it too, sir, if you don't mind.”
Sir Clinton suppressed a yawn with difficulty.
“I don't mind, inspector. Meet us at Neptune's Seat at eleven
o'clock to-morrow morning. It'll be interesting to hear how far wrong
I've gone in some of my guesses; and you can tell me that, since
you've got so much out of these two precious scoundrels to-night.
And now I'll drive you into Lynden Sands—save you the trudge. After
that I really must get to bed.”
Chapter XVII.
The Threads in the Case
“This hasn't been a tidy case, in the strict meaning of the words,”
Sir Clinton mused, as he chose a comfortable spot on Neptune's
Seat and settled down on it. “It's really an omnium gatherum. It
began long before we appeared on the scene; and the inspector has
the facts about the earlier stages, whilst I've nothing better than
guesswork.”
“What we're interested in, chiefly, is what you thought about it at
different stages in the game,” Wendover pointed out. “If you start
with the Peter Hay case and go on from there, you can tell us what
you saw and what we missed. And at the tail end you can give us
your guesses about the earlier stage. The inspector can check them
from the confessions he got.”
Sir Clinton agreed with a gesture, and began without more ado. It
was evident that he was by no means eager to recapitulate, and was
doing so merely out of good nature.
“The Peter Hay case was crystal clear so far as one side was
concerned. It required no marvellous insight to see what had
happened. It wasn't by any possibility a one-man murder. At least
two men must have been on the spot to overpower Peter and tie him
up. They—or at least one of them—was a better-class fellow, or
Peter Hay would have been in his shirt-sleeves instead of having his
jacket on. And the jacket implied that he knew they were coming that
evening, too. Further, the fact that they had amyl nitrite ready in their
pockets is enough to prove two things. They weren't casual
strangers, for they knew about his liability to cerebral congestion.
And they premeditated killing him in certain circumstances. We
worked out pretty definitely the course of events which led to his
death, so I needn't go over that again. I suppose we were right in the
main points, inspector?”
Armadale, primed with the information he had extracted from
Aird, was able to confirm this.
“They used surgical bandages because they hoped to leave no
marks on the skin, I suppose?” Sir Clinton inquired.
“That was the idea, Aird admitted, sir. He thought they'd
succeeded, and he was surprised to find they'd made a mess of it.”
Sir Clinton smiled, apparently at the thought of Aird's
discomfiture.
“Well, if they bungled one side of the affair, they certainly
managed to leave us in the dark about the second side: the motive
behind the murder. It wasn't robbery, obviously. It wasn't bad feeling
—everyone we examined had a good word for Peter Hay. It couldn't
be homicidal mania—not with two of them mixed up in it. That left, so
far as I could see, only one probable motive at the back of the thing.
Either they wanted to force Peter Hay into something he didn't like,
or else he knew something about them which they were afraid of.
“Peter Hay wasn't the blackmailing type, from all we learned
about him. That notion wouldn't hold water for a moment. So, if he
had to be silenced, then the information he had must have been
something he'd come across quite innocently. What it was I couldn't
guess. In fact, the whole idea was very vague in my mind, since
there didn't seem anything definite to support it. That was the stage
I'd reached when we sat in the garden and discussed the thing.
“But there was one thing that seemed to lead on to something
fresh—the silver we found in Peter Hay's drawer. That silver was the
murderers' mistake—the silly thing that gives them away. No
sensible person would ever have tried to throw a suspicion of breach
of trust on Peter Hay. The thing was absurd on the face of it, from
what we learned of his character. But these fellows weren't looking at
it objectively. They would have abused their trust in Hay's shoes; and
naturally they saw nothing outré in faking a petty theft and trying to
throw it on his shoulders. That made me think.
“Why suggest a robbery at Foxhills? For that was what they
evidently intended the police to swallow. As a matter of fact, they'd
have been much better to have left the thing alone. It was a mistake
to fake evidence; it always is. However, they'd done it, and I wanted
to know why they'd done it. So, when the Fordingbridges arrived, we
went off to Foxhills.
“I wasn't surprised to find that the murderers had tried to make
the thing convincing by filling that sack with silver odds and ends. I'd
almost expected something of the sort. But I don't suppose any of us
were misled to the extent the murderers hoped. They thought we'd
be delighted to find confirmation of Peter Hay's dishonesty—stuff
ready packed up for transport. But what one obviously asked oneself
was the question: ‘What's all this meant to cover?’ And the answer
was, so far as I could see: ‘The removal of some inconspicuous
object whose absence won't be noticed in the excitement over the
silver.’
“It wasn't long before I got an inkling of what that inconspicuous
object was. Miss Fordingbridge was able to tell us. But, notice, if
Miss Fordingbridge hadn't happened to come up to Foxhills when we
paid our visit, we'd never have known that the diary was missing.
We'd have missed the main clue in the whole affair. That was where
we were really lucky, and where Aird & Co. had very hard lines.
“There was the diary gone a-missing, anyhow; and the obvious
question to ask was: ‘Cui bono? Who stands to score by its
removal?’
“You heard the story of the missing nephew and the power of
attorney given to Paul Fordingbridge. You could hardly help drawing
the obvious conclusion that here at last was a possible motive
appearing behind the Peter Hay case. The stake on the table was
the assets of the Fordingbridge estate; and that was quite big
enough to make murder worth while, if you happened to have a turn
for that sort of thing.
“But when you come to ask: ‘Cui bono?’ you find that the
problem's like one of those quadratic equations we used to get at
school, where there are two answers and one seems just as good as
the other.
“Suppose the claimant was an impostor, and see where that
leads you. Derek Fordingbridge had spent a lot of time with Peter
Hay in earlier days. The chances were that Peter Hay was the only
man who could give evidence about their joint doings, and that
evidence might form the basis of a damaging examination of the
impostor. Further, the diary would be a priceless thing for an
impostor to lay his hands on. It would supply him with any amount of
irrefutable evidence which he could draw on for his sham
recollections. Clearly enough, if the claim was a fraudulent one, then
the theft of the diary and the silencing of Peter Hay would fit in very
neatly.
“On the other hand, suppose the boot's on the other foot.
Assume that Paul Fordingbridge had some very strong reason for
wishing to retain control over the funds, and see where that leads
you. Remember that he showed no desire whatever to investigate
this claim. He simply denied straight off that the claimant was his
nephew, without waiting for any evidence on the point. That seemed
to me a curious attitude in a trustee; perhaps it struck you in the
same way. And he appeared to be very little put out by Peter Hay's
death, if you remember—treated it very much as a matter of course.
It doesn't take much thinking to see that what holds good for a
fraudulent claimant would hold good for a fraudulent trustee also.
The diary and Peter Hay would be two weak spots for him too.
They'd help a genuine nephew to establish an almost irrefutable
case if he could pass tests applied both by the diary and by Peter
Hay's recollections.
“So, whichever way one looked at it, there seemed to be
something to be said. And, consequently, I got no further at that
stage than being able to say that three things were possible. First,
the claimant might be a fraud, and Paul Fordingbridge merely an
obstinate old beggar. Second, the claimant might be genuine, and
Paul might be a dishonest trustee. Or, third, both the claimant and
Paul might be wrong 'uns.
“Miss Fordingbridge had known her nephew intimately, and she
had identified him straight off, it's true. But we've heard of that kind of
thing before. You remember how Roger Tichborne's mother identified
Arthur Orton as her son, and stuck to it through thick and thin.
Hallucinations of that sort do occur. And one couldn't help noticing
Miss Fordingbridge's talk about spiritualism and so forth, all tending
to show that she had a sort of fixed idea that her nephew would turn
up sooner or later. That discounted the value of her identification a
good deal, but it didn't discredit it completely, of course.
“Now go back a stage. The thing was a two-man job at Peter
Hay's. Therefore, whether the claimant or Fordingbridge was our
man, we had to find a second fellow for the accomplice's part. The
claimant we knew nothing about at that stage; and I proposed to look
into his affairs later. If Paul Fordingbridge was one of the murderers,
on the other hand, then, who was his accomplice? ‘Cui bono?’ again.
If the claimant could be kept out of the succession, who was next on
the list? Stanley Fleetwood's wife.”
Careless of the inspector's feelings, Wendover broke out at this
point:
“You won't persuade me you were such an ass, Clinton, as to
suppose that young Fleetwood helped in a murder merely for the
sake of cash or any other reason?”
“It wasn't my business to make pets of anyone, and exclude them
from suspicion merely because I liked them in private life, squire.
Many murderers are most amiable persons—Crippen, for example.
‘A fair field and no favour’ is the only motto for a conscientious
detective.
“Before we had time to delve further into the Peter Hay case,
however, the Staveley murder occurred. There's no need to go into
the whole business; it's fresh in your minds; but I'll tell you the main
points that struck me when we'd finished our examination of the
scene of the murder.
“First, Staveley had banged his wrist and stopped his watch at
11.19. But, of course, that didn't prove he'd been killed at that
moment. Second, his clothes were wet under his rainproof; and, he'd
been shot through both rainproof and jacket, it must have rained
before he was shot. Third, since the car-tracks had gone back on a
dry road for a while before the rain came on and made them clearer,
Staveley was killed after the car had gone off; and the people in the
car weren't mixed up in the actual killing. Fourth, there was only one
cartridge-case to be seen—the one on the rock. There was no
cartridge-case at the groyne when I searched the place. Besides,
that track at the groyne belonged to the man in the car, and he was
cleared completely by the rain question. If I was right in my
inferences, then the murder must have been committed by one of
three people: the woman with the neat shoes, Billingford, or
someone who had left no tracks on the sand.
“The letter we found in Staveley's pocket showed the business;
and you worked up the case against them that Mrs. Fleetwood had
been to meet him on the previous night at the rock; and, as she was
acting in conjunction with a man, there wasn't much trouble in
inferring that young Fleetwood might have been on the spot also, as
soon as we heard that the Fleetwoods' car had been out all night:
You, inspector, jumped to the conclusion that the Fleetwoods were at
the back of the business; and you worked up the case against them
very convincingly. But, as I told you at the start, the case wasn't
sound. I wanted all the data I could get, of course, so I didn't
discourage you too much; and you eventually dragged out a lot of
interesting material about the events of the night.
“Meanwhile, we'd come to a blank end with the Fleetwoods and
had turned to Billingford. My impression was that he seemed
genuinely surprised by the news of Staveley's murder; but he might
have been acting, for all one could tell. What we did get out of him
was the general impression that Flatt's cottage was inhabited by a
gang of rogues. How many were there? Three, if one took Billingford
at face value; four, if one believed the story the claimant blurted out
when we questioned him at the cottage.
“Anyone could see that the fourth man was a dark horse. He
might be the murderer whom they were shielding, possibly. But there
was another explanation of his disappearance; he might be someone
well known to the local people, and whom it was desirable to keep
under cover. How would that fit in with things? Suppose the claimant
was an impostor; he wouldn't be very anxious to meet the villagers
more than he could help, for fear of dropping on someone who might
trap him and expose him. The less he saw of his neighbors the
better; and his disfigurement gave him a fair excuse for keeping
indoors in the daytime. Staveley was well enough known to the
villagers also; and perhaps he had good reason for not wishing his
presence known. If the fourth man was in the same boat, then none
of them would care to go shopping and so forth, and yet supplies
had to be got daily. Hence it might be convenient to have a man like
Billingford as the nominal host, to act as go-between for them in their
public transactions. That's how it appeared to me. Naturally, I was
curious about the fourth man, and I got you, inspector, to set a watch
and see if he could be recognised.
“That left me with a fair suspicion that these fellows were
hatching some devilment or other at the cottage. Then I noticed the
card-index; and I saw light to some extent. A card-index implies the
need for ready reference. The claimant, if he were a fraud, would
need to cram himself with all the available facts about the doings of
the real Derek Fordingbridge—just as Arthur Orton crammed up all
the facts about Roger Tichborne. And a card-index would be the
handiest repository of all the news they could collect. As you saw for
yourself, squire, that guess of mine was right.
“Assume that state of affairs—I had no certain knowledge then—
and things begin to fall into their places. I've given you my notion of
why Billingford was needed. What about the other three?
“The claimant was obviously needed to represent Derek
Fordingbridge; and he'd been cast for the part on two grounds. First,
his face was so much damaged that no one could swear to his
original appearance. He might quite well have been Derek or anyone
else, so far as that went. Then the loss of his fingers made him
invaluable also, because he couldn't be expected to write like the
real Derek nowadays, with a mutilated hand. All that was wanted in
addition was a good memory to cram up the immense amount of
facts that they needed in order to meet questioning.
“Then there was Staveley. What was he doing in the affair? Well,
obviously, he had a lot of information about the Foxhills people which
he must have picked up while he stayed there with the real Derek on
leave, and also some more facts which he must have learned from
Mrs. Fleetwood from time to time.
“And, finally, there was the fourth man. I suspected that he might
turn out to be a second information-mine; and when I heard the
report you gave me, inspector, about the fishermen and Sapcote
having recognised him as an ex-valet at Foxhills, I felt I was getting
on to fairly sure ground.
“Well, there were four of them to share in the loot if they pulled it
off. But a third's better than a quarter-share any day. If they had
pumped Staveley dry of his information, and had got notes of it all on
that card-index, what further need had they for friend Nicholas?
None whatever.
“And suppose they could involve the Fleetwoods in a murder
case and get them hanged, wouldn't that remove one possible set of
objectors to the claimant dropping into the funds? So I didn't exclude
the possibility that they knew—although they denied it—that
Staveley was going to meet Mrs. Fleetwood at Neptune's Seat that
night. When I say ‘they,’ I really mean the faceless fellow and Aird.
“There was a further long shot possible. I'm not sure if it really
entered into their plans; but I give it you for what it's worth. Suppose
they suggested a walk along the sands to Billingford that evening,
and arranged matters so that he would reach the rock just after the
murder had been committed and they had cleared out. Wouldn't that
have been a tight corner for Master Billingford? With any luck he
might have been hung for the murder, since he'd no evidence but his
own to rely on to prove he wasn't on the spot when the shot was
fired. And then there would be only two of them, instead of four, to
share out the loot if they got it.
“You see now how I was beginning to look at the affair. But I was
considerably worried by the woman with the neat shoe. Her part in
the business would have to be cleared up eventually; but for the
moment I had to put it aside.
“And then our friends made their second blunder—trying to prove
too much, as usual. Friend Cargill came on the scene, innocently
going down to bathe. He sat down on the groyne and proceeded to
dig up a .38 cartridge-case, which he presented to me like an honest
fellow anxious only to help the police. Well, all three of us had been
over that particular bit of sand and had seen no cartridge-case
before he arrived on the scene. Also, as I pointed out to you, squire,
an automatic ejects its cartridge-case sharply and jerks it well behind
you, especially on hard sand where the thing can jump along. It was
obvious that no one could fire a shot from Fleetwood's position at the
groyne and leave his ejected shell lying close under the groyne,
where Cargill assured me he'd kicked it up. So naturally I began to
look at Mr. Cargill with more than common interest; and, as you saw
yourselves, he's got a build rather like the claimant's, so I wondered
if they were related.
“Then our friend Cargill told us his yarn about meeting Derek
Fordingbridge in the war; and off he went to meet his dear old friend.
And later on he volunteered eagerly that he'd had a talk with the
dear old fellow. By that time I was more than a bit suspicious of the
dear old friend; and naturally some of that suspicion spilled over on
to Cargill. If the claimant was an impostor, then the man who
recognised him was a liar; and, as I had no use for aimless liars in a
case of this sort, I inferred that Master Cargill was one of the gang,
posted at the hotel for intelligence purposes—to keep an eye on the
Fordingbridge group. And that cleared up one of the main difficulties
I'd had—namely, how the murderer had known to use a .38
automatic so as to match the Fleetwood pistol. Of course, if you
assume that Cargill had taken the opportunity of rummaging in
Fleetwood's room, or had drawn him into talk about pistols, they
would be sure of their ground on that point. That had been a
troublesome point to me; for I didn't like to stretch coincidence to the
extent of assuming it was mere accident that made the Fleetwood
pistol and the bullet in the body both of the same calibre.
“It remained to check Billingford's story as far as possible, and
you know how the runnel helped us in that. What the facts of the
case proved beyond any reasonable doubt was that at 11.19 p.m.
Billingford was about three-quarters of a mile from Neptune's Seat.
With the sound of the sea in his ears, it's most unlikely that he could
hear a pistol-shot at that distance. And his tracks showed him
walking along quite steadily there. Then, at a point far nearer the
rock—a point that I suppose he may have reached about 11.35 p.m.
—the trail showed that he began to run. Now that fitted in with his
story. At that second point he might quite well have heard a shot
fired on the rock, just as he said he did. He couldn't possibly, on the
facts we established, have reached the rock before about 11.37 or
11.38; and by that time the murder was done and the murderer had
got away.
“By that time I felt fairly sure of my ground; and I got that digging
in the sand started, just on the off-chance that we might get hold of
the shell of the cartridge which really did kill Staveley. It wasn't
absolutely necessary for the case; but, if it turned up, then it would
help to confirm my notions. As it was bound to be below tide-mark,
there was no point in trying to locate its exact position, since it might
have been washed about by the waves in the falling tide after the
shot was fired. So I simply had the whole strip of sand dug up and
dumped down above the high-water mark for future examination.
“Meanwhile, I'd been on the hunt for the dame with the neat shoe
—a No. 4, as you remember. There were several of the guests
wearing shoes of that size; but I picked out Mme. Laurent-
Desrousseaux first of all because she seemed likeliest. Staveley had
served in France; she was a Frenchwoman; she was at Lynden
Sands Hotel, where she most obviously was out of place and knew
no one. It seemed best to find out something about her.
“I talked to her. She was a bit lonely, it seems, and quite ready to
go for a walk now and again. I know my own length of pace; so, by
counting hers in a given distance and comparing with the numbers of
my own, I made a fair guess at her step-measurement. It fitted in
with the prints of the neat shoe on the sands. Gentle treatment did
the rest, as you saw. She told us her story quite honestly, and it
threw a good deal of light on the affair. I inferred from it just what you
yourself inferred later on, squire: the shot fired by Mrs. Fleetwood at
11.19; Staveley's fall on the rock; and the bolt in the car to the hotel.
And, in turn, this checked Billingford's story quite neatly, because he
couldn't have heard the shot at 11.19, being so far away. It was the
second shot, about 11.35, that he heard.
“Now, let's reconstruct what really happened; and remember that,
although it was full moon, it was a cloudy night, and the light was
bad all through. We'll begin with Staveley leaving the cottage. He's in
a bad temper; been drinking as well as playing poker. He gets to the
rock and waits for Mme. Laurent-Desrousseaux. She's late; and his
temper gets worse. She arrives and tells him she wants to get a
divorce arranged. Between his temper and the chance he sees of
making her pay sweetly for the favour she asks he treats her brutally,
and sends her off both pained and angry. Then Mrs. Fleetwood
arrives, and that meeting culminates in the shot she fired at 11.19 by
accident. Then you get the talk between the Fleetwoods at the car,
overheard by Mme. Laurent-Desrousseaux; their drive back to the
hotel; and her hurry to get away from what she thinks is the scene of
a murder. That leaves Staveley on the rock, stunned by his fall; and
Billingford sauntering across the sands towards the runnel. In the
meantime, Aird and the gentleman without a face have got out the
boat belonging to Flatt's cottage and are rowing for Neptune's Seat,
which is just above tide-level.
“Down comes the rain. Staveley gets soaked; and perhaps the
chill revives him. He staggers up and puts on his rainproof. Then in
comes the boat, and they shoot him without having to land on the
rock at all. The ejector jerks the cartridge-case into the water, where
it sinks into the sand and gets covered up by the wash of the waves.
The murderer and his pal row off into the dark. Meanwhile Billingford
has heard the shot, and he, not knowing anything about this little
plot, rushes up—rather pluckily—to see what it's all about, and he
finds the body on the rock. That explains the two shots and the
general chronology of the affair.
“Now, by this time friend Cargill had made his second error. He'd
been keeping his eye on things at the hotel, and he'd got hold of that
envelope which Mme. Laurent-Desrousseaux addressed to Staveley
and dropped in the waste-paper basket. He thought he could give
me a fresh scent to follow up, and he was just going to produce it
when he realised that Mme. Laurent-Desrousseaux herself was with
us; so he suppressed it then and handed it over later on. All it did
was to confirm me in the idea that he was one of the gang.
“Up to that point everything seemed plain sailing. I had, as I
believed—and as it turned out in the end—got the thing cut and dried
against the gang at the cottage. The case wasn't complete; but,
short of getting some direct evidence from one of the two actual
murderers, I didn't see how I was to make it absolutely water-tight.
They'd been a bit too clever for a jury, I feared. And, of course, the
Peter Hay case was getting clearer also, once you could assume
that these fellows would stick at nothing.
“Then, out of the blue, came the shooting of Cargill. That wouldn't
fit in with the rest of the business. Cargill wasn't a pawn like Staveley
and Billingford. He was watching one end of the business for them—
keeping an eye on us for one thing; and, besides, I was becoming
more and more sure that he was a brother of the faceless fellow, and
possibly the brains of the gang. They had a use for him; they
wouldn't shoot him. But, then, who did?
“And at that point I took a long cast back and raked up again a
possibility I'd dismissed at an earlier stage. Suppose that both the
claimant and Paul Fordingbridge were wrong 'uns, what then?
Suppose friend Paul had been at some hanky-panky with the funds
he held in trust for his nephew. Then, whether the claimant was an
impostor or not, it would be very convenient for friend Paul if the
claimant left this vale of tears. And the claimant and Cargill were
much alike in build; and Cargill was shot after leaving the cottage.
There might be something in it. And when I found that friend Paul
carried a pistol in his pocket, and didn't care who knew it, by the look
of his jacket, I began to think furiously.
“I didn't blame Paul for carrying arms. In his position, with that
gang at the cottage in the offing, I think it was a wise precaution; for
he must have known that he was the main stumbling-block in the
claimant's road. But I don't think that he kept within the limits of
precaution. I think he decided to get ahead of them by knocking out
the claimant—and after that he would be able to live in peace as
heretofore.
“However, I never had time to probe that matter any further, for
the next business was the disappearance of friend Paul. I think I
have a fair notion how that was managed.
“They approached him and asked for an interview. He sent the
claimant a scrap of paper:
‘Meet me at the Blowhole to-night at 11 p.m. Come alone.’
The last two words give you the key to friend Paul's feelings
about them. One man he could keep an eye on, and he didn't
propose to have any more present. Of course, they filed that note
and used it again later, as you'll remember.
“Probably the claimant met him at the Blowhole and suggested a
walk over the open sands as a good way of avoiding eavesdroppers.
Paul would feel safer in the open. By the time they reached the old
wreck the claimant would have got him interested, or else his natural
fears would be dissipated. At the hulk the claimant obviously turned,
as though to go back across the sands, and Paul turned with him.
Then, from behind the hull, Aird stole out and did le coup du Père
François.”
“What's that?” Wendover demanded. “You talked a lot about Père
François and Sam Lloyd's ‘Get off the Earth’ puzzle, I remember.”
“If you happen to be in Paris late at night, squire, and a rough-
looking customer asks you the time or begs for a match, you'd better
look out for his friend—le Père François, they call him—who may be
coming up behind you with a long strip of flannel in his hand. While
the first man holds you in talk, the Père François lassoes you with
his flannel rope and pulls the two ends so that it catches your throat.
Then he sinks down suddenly and turns his back to you, slipping the
rope over his shoulder as he turns. This pulls you down back to back
with him; and when he rises to his normal height again, there you are
on his back like a sack on a coal-heaver's back, with your feet off the
ground. The first man then goes through your pockets at his leisure,
and if you choke to death before he's done, so much the worse for
you. You can't struggle with any effect.
“That I suspect, was how they caught friend Paul; and Aird just
carried him on his back to the quicksand and dumped him in. From
Aird's footmarks it was clear he'd been carrying a heavy weight; the
prints were deep and the feet almost parallel after he'd done his
Père François trick. See now what I meant by ‘Get off the Earth’?
Naturally there were no signs of a struggle, since all the struggle was
off the ground. And, of course, they'd take care to wear shoes that
left no clue—common type and largest size. And they got away
either in a boat or by wading along in the water, so as to leave no
tracks. I could see no way to bring the affair home to them. The only
sure method depended on our wringing evidence out of one of them
somehow; and I didn't see how it could be managed just then. Also, I
hadn't much of a case against Cargill beyond suspicion; and I
wanted him too, if it could be managed.
“The next thing was the arrival of the Fordingbridge lawyer; and
from him I learned that we might get on the track of any
malversations by Friend Paul if I went up to London. I wanted to
know definitely where I stood in that matter, because, if I was wrong
there, then the whole latter part of my notions would collapse. So I
made up my mind to go to town.
“But I was very uneasy. Now that Paul Fordingbridge was out of
the road for good, Mrs. Fleetwood was the only person between the
claimant and the cash. If she disappeared in her turn, then Miss
Fordingbridge would have welcomed her long-lost nephew with pure
joy and gratitude for his preservation, and there would have been no
one left alive to object to his coming into Foxhills and the rest.
Therefore, I was inclined to take some steps to see that she came to
no harm while I was away.
“The obvious thing to do would have been to warn her. But that
would have meant giving my case away to the Fleetwoods; and I
don't feel inclined to chuck confidences around if it can be avoided,
as I've pointed out before. Further, the police were not altogether in
good odour with the Fleetwoods; and I wasn't sure if I'd make much
impression on them by a mere warning, with nothing to back it. So I
hit on the notion of putting a man on to watch Mrs. Fleetwood; and,
as a further precaution, I fixed up that code-wire so that if I wanted it
I could have her arrested at a moment's notice, and then she'd be
safe in police hands and out of reach of the gang.
“I went up to London and found, as I'd expected, that friend Paul
had been playing ducks and drakes with all the securities he could
handle without exciting too much suspicion. He seems to have been
speculating right and left, most unsuccessfully. So I'd been right
about his motives, anyhow.
“But I couldn't get out of my mind the risk I was letting that girl
run; and at last—I suppose Miss Fordingbridge would say it was
telepathy or something—I got the wind up completely, and wired to
have her arrested. After that I felt safer.
“As you know, they'd been too quick for me. They fished out the
note that Paul Fordingbridge wrote to the claimant and they sent it to
her as if it came from Paul himself, after altering the hour on it. She
thought her uncle was in trouble; went to help him; dodged the
constable; and fell straight into the trap they'd set for her. You know
all the rest. And probably by now you understand why I was quite
content to let Mr. Aird have his full dose in the funnel of the souffleur.
There's nothing like a confession for convincing a jury, and I meant
him to hang if it could be managed. I didn't want to run any risk of his
getting off merely because it was all circumstantial evidence.”
“Thanks,” said Wendover, seeing that the chief constable had
finished his outline. “To quote from that favourite detective story of
yours:
In one moment I've seen what has hitherto been

Enveloped in absolute mystery.
There's just one point I'd like to hear you on. What about
Staveley's resuscitation after his being killed in the war? Did you get
to the bottom of that by any chance?”
Sir Clinton hesitated a little before answering.
“I don't much care about pure guesswork, squire; but, if you'll
take it as that, then I don't mind saying what I think. Suppose that is
what happened. Staveley and Derek Fordingbridge went into action
together; and Staveley was under a cloud at the time. He'd probably
had enough of the war, and was looking for a way out. Derek
Fordingbridge gets killed in that battle, and is probably badly
damaged in the process—made unrecognisable we may suppose.
Staveley sees him killed, and grasps at the chance offered. He takes
Derek's identity disc off the body and leaves his own instead.
Probably he takes the contents of the pockets too, and puts his own
papers into the dead man's pockets. They were friends; and, if
anyone saw him at work, he'd have his excuse ready. No one would
think he was robbing the dead. Then he goes on—and simply hands
himself over to the enemy. He's a prisoner of war—under Derek's
name.
“He manages to escape, and the escape is put down to Derek's
credit. But, of course, Derek never turns up again; and naturally
people suppose that he must have died of exposure in his last
attempt, or been shot at the frontier, or something of that sort.
Meanwhile Staveley, once out of Germany, drops his borrowed
identity, probably changes his name, and disappears. I suspect he
was in very hot water with the military authorities, and was only too
glad of the chance to vanish for good.
“After the war, he evidently got in amongst a queer gang, and
lived as best he could. Billingford's evidence points to that. And
somewhere among this shoal of queer fish he swam up against our
friend Cargill. My reading of the thing is that somehow Staveley gave
away—perhaps in his cups—something of what I've given you as my
guess; and Cargill, remembering his disfigured brother, saw a grand
scheme to be worked by putting forward his brother as claimant to
the Foxhills property.
“It wasn't half so wild a plan as the Tichborne business, and you
know how that panned out at the start. So the three of them set to
work to see the thing through. Staveley, I suspect, got hold of Aird,
who had invaluable information about all the affairs at Foxhills in the
old days. Then they went to work systematically with their card-index
and noted down everything that Aird and Staveley could remember
which would bear on the case.
“That accounts for the delay in the claimant turning up. It
probably was quite recently that Staveley fell in with Cargill. And
evidently the delay points to the fact that Staveley wasn't the
originator of the notion, else he'd have got to work much earlier. It
was only when he fell in with Cargill, who had a brother suitable to
play the part of the claimant, that anything could be done. Then they
must have spent some time in unearthing Aird.
“Well, at last they're ready. They come down to Lynden Sands
with their card-index handy. Now, the claimant doesn't want to
appear in public more than he can help, for every stranger is a
possible danger to him. He might fail to recognise some old friend,
and the fat might be in the fire. Nor does Staveley want to show
himself; for his presence might suggest the source of the claimant's
information. Aird's in the same position. And when they learn that the
Fordingbridges are at the hotel, Cargill is detached there to keep an
eye on them. Thus they need a go-between; and Billingford is
brought down to serve that purpose. Also, as soon as the claimant
makes his first move there will be sure to be a lot of gossip in the
village, anecdotes of the claimant's history floating round, and so
forth; and Billingford will be able to pick them up and report them to
the rest of the gang. They'd have been safer to leave Staveley and

Full Ebook of Advances in Data Science and Computing Technology Suman Ghosal Online PDF All Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Full Ebook of Advances in Data Science and Computing Technology Suman Ghosal Online PDF All Chapter

Uploaded by

Copyright:

Available Formats

Advances in Data Science and

Computing Technology Suman Ghosal

Advances in Data Science and Computing Technology:

Advances in Digital Science ICADS 2021 Advances in

Advances in Data Science 1st Edition Ilke Demir

Advanced Soft Computing Techniques in Data Science, IoT

Sustainable Production and Applications of Waterborne

Introduction to Quantum Computing (River Publishers

The Farinograph Handbook : Advances in Technology,

Frontiers of Data and Knowledge Management for

ADVANCES IN DATA SCIENCE AND

© 2023 by Apple Academic Press, Inc.

Library and Archives Canada Cataloguing in Publication

CIP data on file with US Library of Congress

ISBN: 978-1-77463-997-9 (hbk)

Most real-world search and optimization problems naturally involve multiple

the use of analytical methods that examine trade-offs; consider multiple

• Mathematical programming in MCDM

Features of the volumes will include recent trends, model extensions,

CURRENT & FORTHCOMING BOOKS IN THE SERIES

Suman Ghosal, PhD

Amitava Choudhury, PhD

Vikram Kumar Saxena, MTech

Arindam Biswas, PhD

Prasenjit Chatterjee, PhD

Part I: Recent Application of Deep Learning and ANN Paradigm ................... 1

2. An Efficient and Optimized Deep Neural Network Architecture

3. Design and Development of a Driving Assistance and

4. Prediction of the Positive Cases of COVID-19 for Individual States

5. Deep Residual Neural Network Based Multi-Class Image

6. Improved TDOA-Based Node Localization in WSN:

8. A Review of the Revolutionizing Role of Big Data in Retail Industry .... 95

10. Load-Balancing Algorithm with Overall Tasks....................................... 119

11. Implementation of Load Balancing of Matrix Problem with the

Part IV: Advances in Embedded System-Based Applications ....................... 143

13. Design and Modifications of Eyedrop Bottle Holder Device:

14. Agro-Nutrition Alert (ANA)...................................................................... 187

Part V: Optimization Technique Using MATLAB Platform.......................... 195

16. High Efficiency High Gain SEPIC-Buck Boost Converter Based

17. A Classy Fuzzy MPPT Controller for Standalone PV System............... 237

18. Active and Reactive Power Control in Variable Wind Turbine

20. Strength-Based Novel Technique for Malicious Nodes Isolation........... 289

21. Securing the Information Using Combined Method .............................. 305

22. Implementation of DNA Cryptography in IoT Using Chinese

23. Robust Technique to Overcome Thwarting Communication

Part VII: Miscellaneous Topics in Computing ................................................ 347

25. Comparative Analysis of Storey Drift of a G+9 RC-Structured

26. A Proposal for Disinfection of Contaminated Subject Area

Index .................................................................................................................... 397

Madhu Agarwal Agnihotri

Pushpanjali Singh Bisht

Aditya Narayan Chakraborty

Palash Kumar Kundu

Bijoy Kumar Mandal

Ranjan Kumar Mondal

Mausumi Das Nath

Mahua Nandy Pal

Rahul Kumar Prasad

Ananya Satpati (Das)

Vikram Kumar Saxena

Rahul Kumar Singh

Arunava Kabiraj Thakur

Parag Kumar Guha Thakurta

point of effective deep convolutional neural network (CNN) vessel classifi