SpringerBriefs in Computer Science

Series Editors

Stan ZdonikPeng NingShashi ShekharJonathan KatzXindong WuLakhmi C. JainDavid PaduaXuemin ShenBorko FurhtV. S. SubrahmanianMartial HebertKatsushi IkeuchiBruno Siciliano

For further volumes:http://www.springer.com/series/10028

Murat _Ilsever • Cem Ünsalan

Two-Dimensional ChangeDetection Methods

Remote Sensing Applications

123

Murat _IlseverDepartment of Computer EngineeringYeditepe UniversityAgustos Yerles�imi 2634755 KayisdagiIstanbulTurkey

Cem ÜnsalanElectrical and Electronics EngineeringYeditepe UniversityAgustos Yerles�imi 2634755 KayisdagiIstanbulTurkey

ISSN 2191-5768 ISSN 2191-5776 (electronic)ISBN 978-1-4471-4254-6 ISBN 978-1-4471-4255-3 (eBook)DOI 10.1007/978-1-4471-4255-3Springer London Heidelberg New York Dordrecht

Library of Congress Control Number: 2012940953

� Cem Ünsalan 2012This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part ofthe material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission orinformation storage and retrieval, electronic adaptation, computer software, or by similar or dissimilarmethodology now known or hereafter developed. Exempted from this legal reservation are briefexcerpts in connection with reviews or scholarly analysis or material supplied specifically for thepurpose of being entered and executed on a computer system, for exclusive use by the purchaser of thework. Duplication of this publication or parts thereof is permitted only under the provisions ofthe Copyright Law of the Publisher’s location, in its current version, and permission for use must alwaysbe obtained from Springer. Permissions for use may be obtained through RightsLink at the CopyrightClearance Center. Violations are liable to prosecution under the respective Copyright Law.The use of general descriptive names, registered names, trademarks, service marks, etc. in thispublication does not imply, even in the absence of a specific statement, that such names are exemptfrom the relevant protective laws and regulations and therefore free for general use.While the advice and information in this book are believed to be true and accurate at the date ofpublication, neither the authors nor the editors nor the publisher can accept any legal responsibility forany errors or omissions that may be made. The publisher makes no warranty, express or implied, withrespect to the material contained herein.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Preface

Sequential images captured from a region may be used to detect changes there.This technique may be used in different fields such as video surveillance, medicalimaging, and remote sensing. Especially in remote sensing, change detection isused in land use and cover analysis, forest or vegetation inspection, and floodmonitoring. Although manual change detection is an option, the time required forit can be prohibitive. It is also highly subjective depending on the expertise of theinspector. Hence, the need for automated methods for such analysis tasks emerged.This book is about such change detection methods from satellite images. Our focusis on changes in urban regions. The layout of the book is as follows.

We start with a brief review of change detection methods specialized for remotesensing applications. While the first Earth observation satellites were equippedwith 30-100 m resolution sensors; modern ones can capture images up to 0.5 mresolution. This also led to the evolution of change detection methods for satelliteimages. Early methods were generally pixel based. As the detail in the imageincreased, more sophisticated approaches emerged (such as feature based methods)for change detection.

Next, we consider pixel based change detection.We summarize well-knownmethods such as: image differencing, image ratioing, image regression, and changevector analysis. We introduce median filtering based background subtraction forsatellite images. We also propose a novel pixel based change detection methodbased on fuzzy logic.

To benefit from color and multispectral information, we explore severalmethods such as PCA, KTT, vegetation index differencing, time dependentvegetation indices, and color invariants. Since these methods depend on a linear ora nonlinear color space transformation, we labeled them as such. Naturally, theycan only be applied to the dataset having color or multispectral information.

We also considered texture based descriptors for change detection. Here, webenefit from the gray level co-occurrence matrix. We extracted four texturedescriptors from it to be used for change detection. We also benefit from entropy tosummarize the texture.

v

Different from previous approaches, we introduced a change detection frame-work using structure information. Here, we extract the structure in an image byedge detection, gradient magnitude based support regions, matched filtering, andlocal features. Graph formalism also helped us to summarize the structure in theimage.

Finally, we introduced fusion of change detection methods to improve theperformance. Since different change detection methods summarize the changeinformation in different ways, they can be fused to get a better performance.Therefore, we considered the decision level fusion based on binary logic. We alsodeveloped a fusion method based on association.

We statistically evaluated the performance of the mentioned change detectionmethods. On a large dataset, we obtained very promising results. Especially, thechange detection performance after fusion is noteworthy.

The brief summary above indicates that this book may be useful for automatedchange detection studies. It summarizes and evaluates the existing methods onchange detection. It also proposes several novel methods for satellite image basedchange detection. Therefore, the interested reader may benefit from both categoriesto solve his or her research problems.

Istanbul, Turkey, May 2012 Murat _IlseverCem Ünsalan

vi Preface

Acknowledgments

The authors gratefully acknowledge the financial support of The Scientific andTechnological Research Council of Turkey (TUBITAK), in the framing andexecution of this work through project number 110E302.

vii

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Literature Review on Change Detection in Satellite Images . . . . 11.2 Layout of the Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Pixel-Based Change Detection Methods. . . . . . . . . . . . . . . . . . . . . 72.1 Image Differencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.2 Image Rationing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.3 Image Regression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.4 Change Vector Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.5 Median Filtering-Based Background Formation. . . . . . . . . . . . . 162.6 Pixelwise Fuzzy XOR Operator. . . . . . . . . . . . . . . . . . . . . . . . 17References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3 Transformation-Based Change Detection Methods . . . . . . . . . . . . 233.1 Principal Component Analysis . . . . . . . . . . . . . . . . . . . . . . . . 233.2 Kauth-Thomas Transformation . . . . . . . . . . . . . . . . . . . . . . . . 263.3 Vegetation Index Differencing . . . . . . . . . . . . . . . . . . . . . . . . 273.4 Time-Dependent Vegetation Indices. . . . . . . . . . . . . . . . . . . . . 293.5 Color Invariants. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

4 Texture Analysis Based Change Detection Methods. . . . . . . . . . . . 354.1 Gray Level Co-occurrence Matrix . . . . . . . . . . . . . . . . . . . . . . 354.2 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5 Structure-Based Change Detection Methods . . . . . . . . . . . . . . . . . 415.1 Edge Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415.2 Gradient-Magnitude-Based Support Regions . . . . . . . . . . . . . . . 42

ix

5.3 Matched Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425.4 Mean Shift Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435.5 Local Features. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455.6 Graph Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465.7 Shadow Information. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

6 Fusion of Change Detection Methods . . . . . . . . . . . . . . . . . . . . . . 536.1 Fusion Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536.2 Category Level Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546.3 Inter-Category Level Fusion . . . . . . . . . . . . . . . . . . . . . . . . . . 54

7 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577.1 The Data Set. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577.2 Performance Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

7.2.1 Pixel-Based Change Detection Methods. . . . . . . . . . . . . 617.2.2 Transformation Based Change Detection Methods . . . . . 627.2.3 Texture-Based Change Detection Methods . . . . . . . . . . . 627.2.4 Comparison of Thresholding Algorithms . . . . . . . . . . . . 637.2.5 Structure-Based Change Detection Methods . . . . . . . . . . 657.2.6 Fusion of Change Detection Methods . . . . . . . . . . . . . . 68

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

8 Final Comments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

x Contents