(Download PDF) Fuzzy Machine Learning Algorithms For Remote Sensing Image Classification 1St Edition Anil Kumar Online Ebook All Chapter PDF
(Download PDF) Fuzzy Machine Learning Algorithms For Remote Sensing Image Classification 1St Edition Anil Kumar Online Ebook All Chapter PDF
(Download PDF) Fuzzy Machine Learning Algorithms For Remote Sensing Image Classification 1St Edition Anil Kumar Online Ebook All Chapter PDF
https://textbookfull.com/product/image-analysis-classification-
and-change-detection-in-remote-sensing-with-algorithms-for-
python-fourth-edition-canty/
https://textbookfull.com/product/image-analysis-classification-
and-change-detection-in-remote-sensing-with-algorithms-for-
python-4th-edition-morton-john-canty/
https://textbookfull.com/product/machine-learning-algorithms-for-
industrial-applications-santosh-kumar-das/
https://textbookfull.com/product/image-processing-and-gis-for-
remote-sensing-techniques-and-applications-second-edition-liu/
Multisensor Image Fusion and Data Mining for
Environmental Remote Sensing First Edition Bai
https://textbookfull.com/product/multisensor-image-fusion-and-
data-mining-for-environmental-remote-sensing-first-edition-bai/
https://textbookfull.com/product/remote-sensing-and-giscience-
challenges-and-future-directions-pavan-kumar/
https://textbookfull.com/product/deep-learning-for-remote-
sensing-images-with-open-source-software-1st-edition-remi-
cresson/
https://textbookfull.com/product/hyperspectral-imaging-remote-
sensing-physics-sensors-and-algorithms-1st-edition-dimitris-g-
manolakis/
https://textbookfull.com/product/remote-sensing-of-vegetation-
hyperspectral-indices-and-image-classifications-for-agriculture-
and-vegetation-second-edition-huete/
Fuzzy Machine Learning
Algorithms for Remote
Sensing Image Classifcation
When on a joke we laugh once....
Than why we cry number of times on a problem....
Life is Fuzzy....
Use Fuzzy Logic....
....Remain Crisp....
Fuzzy Machine Learning
Algorithms for Remote
Sensing Image Classifcation
Anil Kumar
Priyadarshi Upadhyay
A. Senthil Kumar
First edition published 2020
by CRC Press
6000 Broken Sound Parkway NW, Suite 300, Boca Raton, FL 33487-2742
Reasonable efforts have been made to publish reliable data and information, but the author and publisher can-
not assume responsibility for the validity of all materials or the consequences of their use. The authors 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 [email protected]
Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used
only for identification and explanation without intent to infringe.
Typeset in Palatino
by Lumina Datamatics Limited
iv
Remember........
Life Once........
Birth Once........
Death Once........
To
Gurus who guided us
Students who worked with us
&
Readers of this book
Depth Sky
Stability Sea
Contents
Foreword ...................................................................................................................xi
Preface...................................................................................................................... xv
Our Gratitude with three R’s...................................................................................xix
Authors ....................................................................................................................xxi
List of Abbreviations............................................................................................ xxiii
Chapter 2 Ground Truth Data for Remote Sensing Image Classif cation............. 9
2.1 Introduction ...............................................................................9
2.2 Creation of Training Data........................................................ 11
2.3 Criteria for Ground Truth Data ............................................... 12
2.4 Training Data in Machine Learning........................................ 12
2.5 Validation Dataset ................................................................... 14
2.6 Testing Dataset ........................................................................ 14
2.7 Summary ................................................................................. 15
Bibliography....................................................................................... 15
vii
viii Contents
Appendix: A1, SMIC: Sub-Pixel Multi-Spectral Image Classifer P ackage ....... 131
Appendix: A2, Case Studies from SMIC Package............................................ 141
Index .....................................................................................................................185
Foreword
We are living in a rapidly changing world. Buildings serve multiple purposes, agri-
cultural parcels contain various crops, and forests are reduced in size and compo-
sition. Spatial information on land use and land cover is coming to us in various
ways, and in particular remote sensing images are easily available for that purpose.
Their large resources, being really big data, can be freely assessed, but extracting the
essential relevant information requires solid and well developed machine learning
methods. In addition, we are aware that the world outside is often largely uncertain,
or fuzzy. It requires fuzzy machine learning algorithms to respect this uncertainty
after classif cation.
This book contains a comprehensive overview of the latest developments in fuzzy
methods for remote sensing image classif cation. Dr. Anil Kumar together with
his colleagues Dr. Priyadarshi Upadhyay and Dr. A. Senthil Kumar have made a
remarkable effort to expand and test the available machine learning methods for
fuzzy classif cation. As outstanding experts in their f eld, they have made the book at
a high scientif c standard by providing concepts and equations applied to a range of
practical studies. In an appendix the book contains the algorithms that are needed in
the era of big data for an eff cient classif cation of fuzzy objects.
At this place, I can complement the authors with their achievement and can hap-
pily recommend the book to all who are interested in identifying fuzzy information
from space, in short: to classify fuzzy objects from remote sensing images.
This book covers remote sensing image classif cation from multi-sensor and for
different topographic and environmental conditions. Modern machine learn-
ing approaches are used for classif cation into meaningful classes. Traditional
approaches, e.g., statistical and fuzzy logic classif ers are covered with wide spec-
trum. In addition, modern classif cation approaches based on convolutional neural
networks (CNN) with or without long short term memory (LSTM) layers to incorpo-
rate recurrent neural network (RNN) f avor are examined with model identif cation
obtained by deep leaning methods. The book is a fresh presentation of the subject; it
suits established researchers and graduate students as well.
xi
xii Foreword
Many titles on machine learning are available in the market that cover the f eld in an
excellent manner, like Simon Haykin’s Neural Networks and Learning Machines and
Chris Bishop’s Pattern Recognition and Machine Learning, among others. However,
these titles cover the probabilistic approaches to dealing with the problems of clas-
sif cation and regression, while fuzzy set theory based literature that is increasing
in volume is not given its due place in the machine learning space. It is here that
this book by Drs. Anil Kumar, Priyadarshi Upadhyay, and A. Senthil Kumar f lls
an important gap and provides ample complementary material that will immensely
benef t the Master’s/Doctoral student community and faculty members/researchers.
The material like possibility theory based c-means (PCM) and its variants for clus-
tering, hybrid classif ers, temporal data processing and analysis using fuzzy approaches,
fuzzy error matrix, etc. are all now available at one place treated in a unif ed manner
by the authors for the benef t of the reader who would have been looking into multiple
books/journals for a coverage of these topics. Given the excellent contributions made
by the authors to the development of remotely sensed image analysis, the applications
of the machine learning techniques in a fuzzy set theoretic framework covered in this
title will provide useful insights to the reader that are not easily found elsewhere.
The authors introduce machine learning in the frst chapter, followed by ground
truth collection in the second chapter. Again, with the exception of Dr. Russell
Congalton’s book on Accuracy Assessment, the systematic approach to ground truth
collection for training the classifer and to test and validate the results covered in
this chapter is rarely found elsewhere in detail. Chapter 3 deals with programmed
classifers like FCM, PCM, and others, while Chapter 4 deals with learning based
classifers like artifcial neural network (ANN), convolutional neural network (CNN),
recurrent neural network (RNN), long short term memory (LSTM), deep learning
(DL) principles, etc. Chapter 5 deals with hybrid classifers where fuzzy set theoretic
principles are combined with the learning classifers.
Accuracy values are routinely reported even in high prof le journal publications
merely as overall accuracy, kappa coeff cient, or user/producer accuracies. There are
several other important ways to report accuracy such as fuzzy error matrix and
related operators, entropy method, correlation coeff cient, receiver operating charac-
teristics (ROC), and a few others. The icing on the cake is the software tool sub-pixel
multi-spectral image classif er that has built-in many of the algorithms discussed in
the book. This will help the reader explore the algorithms using different datasets
and make comparisons with techniques implemented in commercial software pack-
ages and in public archives like git-hub.
Overall, this book is a very welcome addition to the machine learning literature,
and I recommend that it is seen on the bookshelf of every practitioner/student. Given
the expertise of the authors in remote sensing, it is expected that the practitioners of
remote sensing related image/data analysis will highly benef t from using this book,
but the tools and techniques presented in the book are equally applicable in other
domains, and therefore benef cial to students and researchers from those domains.
Information on natural resources is vital for the growth and sustenance of human
civilization. From time immemorial, this task has been done through very crude
methods, such as by walking through an area to data collection through low orbiting
Earth Observation (EO) satellite data. Since 1972, satellite data has provided a con-
tinuous time series data of the whole world. With the passage of time, the advent of
computer and its fast computing characteristics, remote sensing data and its analysis
has undergone tremendous change and improvement. The f rst EO sensors provided
data at a very coarse resolution where the pixel size was nearly equal to the size of
a football f eld, whereas in this modern era, the minimum size of the pixel is close
to 30 cm.
Earlier, it was thought that f ner the resolution of satellite data better would be
the information. Every advancement has both its advantages and disadvantages. The
interpretation and analysis procedure required an improvement. From simple sta-
tistical methods, the trend shifted towards incorporation of the human mind and its
thinking process, i.e., artif cial neural network to deep learning and now to machine
learning and ensemble classif ers. The need for learning has become important.
More information means more knowledge and more knowledge means better under-
standing. Interpretation approaches have changed. Thousands of human hours have
been spent to analyze data with different perspective. The ocean of knowledge is get-
ting bigger and it has also become deeper. The remote sensing analysts are looking
forward to a scenario where as soon as the data arrives, it is analyzed and presented
in the desired form to the user community – an ideal concept. To achieve such a
state, it requires high and eff cient algorithms to understand the data, correlate to the
existing knowledge, and if possible further update the knowledge and improve the
learning.
This book, Fuzzy Machine Learning Algorithms for Remote Sensing Image
Classif cation being published by CRC press, is one giant leap in learning and under-
standing analysis, in particular, satellite data. The method of introduction to different
approaches ref ects the fundamental knowledge of the authors. Each topic has been
dealt with utmost care keeping in mind the requirements of the user community. The
journey of SMIC has been coolly charted and maneuvered by Dr. Anil Kumar’s own
career and his constant urge to improve and accommodate new approaches has made
it into a power tool. It is time that this software is made available to the user com-
munity. Dr. Priyadarshi Upadhyay has been a constant learner in his area of expert
and with one of his Guru Dr. Anil Kumar, has embarked on a beautiful journey for
quest of knowledge.
I am extremely delighted to see that the two of my rose buds, Dr. Anil Kumar
and Dr. Priyadarshi Upadhyay from my garden have now fully bloomed and the
fragrance of the same has started to spread worldwide. My best wishes to both of
them and may the beacon of knowledge glow brighter in their hands. Dr. A. Senthil
Kumar is a well-known person in the area of image processing and my whatsoever
interaction with him has always left me more knowledgeable.
Congratulations to all the authors and the book for their efforts and hard work.
xiv Foreword
Thanks to CRC Press for having boosted the morale of the authors by accepting
to publish their work.
GOD bless all of humanity
Dr. Sanjay Kumar Ghosh
Professor and Head
Department of Civil Engineering
Indian Institute of Technology Roorkee
Roorkee, India
‘ask not what your country can do for you, ask what you can do for your
country’ ....John F. Kennedy
Preface
With growing utilization of satellite based earth observation data in environmental
and socio-economic management applications, the remote sensing users community
faces an uphill task of learning a variety of image classif cation methods reported
in the literature. Unfortunately, description of these classif cation methods has been
found suited to graduate students with computer science or information technology
background and not really well suited to a large user community with diverse back-
grounds. Moreover, the methods described are largely based on simulated or outdoor
images and do not address the complexity and high data handling requirements,
typically of satellite images.
The purpose of this book is to cover state of the art image classifcation tech-
niques in discrimination of earth objects from remote sensing data. In particular,
the emphasis is on advances in fuzzy based learning methods for preparing land
cover classifcation maps. The primary emphasis in this book is to provide practi-
cal experience to the usage of the technology such as preparation of ground truth
data, mono/bi-sensor temporal data processing to study specifc crop/vegetation
mapping, crop insurance, forest fre mapping, stubble burning, and post disaster
damage analysis while incorporating temporal spectral indices databases generated
using traditional or class based sensor independent approaches. This book covers
from basic methods to the latest ones – fuzzy based machine learning approaches
such as convolutional neural networks, recurrent neural networks, etc. – in detail for
preparing land cover maps. All fuzzy based machine learning algorithms covered in
this book have been included in an in-house developed tool called SMIC: sub-pixel
multi-spectral image classifer .
Among the various end products derived from this information, land use/land
cover maps have a great impact upon monitoring, planning, and development of
resources in individual countries as well as for a region shared by multiple coun-
tries. In recent years, the large volumes of remotely sensed data acquired by global
space agencies collected through various earth observation sensors with wide
application areas are easily available on a global scale. One of the applications
of remote sensing data is thematic map preparation. To achieve this, users face
challenges like handling mixed pixels, removing isolated pixels or pepper and salt
noise, as well as incorporating temporal information for specif c class identif ca-
tion. Advancement in machine learning enables an eff cient and effective way of
extracting geospatial information from remote sensing data. Further, this book
will also discuss about some unique applications of specif c class mapping such as
a specif c crop, disaster effected area mapping, etc. with the use of temporal multi-
sensor remote sensing data.
Application of fuzzy machine learning algorithms can provide highly realis-
tic classif cation results while mapping specif c classes of interest. These machine
learning algorithms can handle mixed pixel problems through a soft computing
approach, non-linear modeling of classes while applying the kernel concept and
xv
xvi Preface
identif cation of specif c classes of interest through the learning algorithm as well
as possibilistic based fuzzy classif ers. Independently, the fuzzy classif ers have suc-
ceeded in addressing the mixed pixel problem; however, these are not able to resolve
the isolated pixel problem, which is then addressed by providing spatial contextual
information to the classif er. The neighborhood pixel information in the form of con-
textual or local information can be incorporated in fuzzy machine learning algo-
rithms to handle noisy pixels as unclassif ed pixels. Markov random f eld (MRF) and
adaptive local information have been identif ed as robust techniques to model spatial
contextual information.
Chapters in the book are arranged in a manner to explain use of machine learn-
ing algorithms for image classif cation in a systematic manner, to be understood by
the professionals and students conveniently. At the end of each chapter, a selected
list of references is also given. Chapter 1 provides the introduction on the basics of
machine learning algorithms and pattern recognition as well as their capabilities for
earth observation. Chapter 2 deals with the importance and necessary ground truth
data for training and validation purposes.
Since this book aims at realizing fuzzy based classif ers like fuzzy c-means
(FCM), possibilistic c-means (PCM), noise clustering (NC), as well as hybrid
fuzzy classif ers by incorporating spatial contextual information into basic fuzzy
based classif er objective function, the middle chapters of the book describe
various types of classif cation algorithms such as fuzzy classif ers, learning based
classif ers, and hybrid fuzzy classif ers. Chapter 3 deals with an introduction about
clustering algorithms and various types of fuzzy and noise clustering based clas-
sif ers. Chapter 4 offers details on the neural and learning based classif ers. It also
includes the upcoming machine learning algorithms, like convolutional neural net-
work (CNN), recurrent neural networks (RNN), long short term memory (LSTM),
and deep learning which can generate much better classif ed data from remote
sensing images.
Chapter 5 covers various hybrid classif ers generated using the hybridization of
methods such as entropy, similarity and dissimilarity, kernel, and contextual with
base fuzzy based classif ers. The contextual information has been included through
MRF as well as using some local convolution methods.
The classif cation algorithms discussed in this book are supervised classif ers,
therefore exposure about collection of manual training data as well as region grow-
ing methods with large number of similarity as well as dissimilarity methods have
been covered. The advantage of region growing methods on training data is that
the homogeneity in training data increases. The importance of hard as well as soft
fraction outputs has also been explained.
Chapter 6 of the book is devoted to use of the multi-temporal data along with
the fuzzy based classifers, as well as deals with the specifc land cover extractions.
Chapter 7 is concerned with the assessment of accuracy methods for classifcation.
Both relative and absolute methods of accuracy have been discussed. The assess-
ment methods of soft classifed outputs through fuzzy error matrix (FERM) with
single as well as composite operator have been included as a relative measure of
accuracy. The entropy, correlation coeffcient, RMSE, and ROC are absolute meth-
ods of accuracy.
2 Fuzzy Machine Learning Algorithms