Change Detection and Image Time Series Analysis 2: Supervised Methods

ISBN-13: 9781789450576 / Angielski / Twarda / 2021 / 272 str.

Atto, Abdourrahmane M.

Spis treści

Preface ixAbdourrahmane M. ATTO, Francesca BOVOLO and Lorenzo BRUZZONEList of Notations xiiiChapter 1. Hierarchical Markov Random Fields for High Resolution Land Cover Classification of Multisensor and Multiresolution Image Time Series 1Ihsen HEDHLI, Gabriele MOSER, Sebastiano B. SERPICO and Josiane ZERUBIA1.1. Introduction 11.1.1. The role of multisensor data in time series classification 11.1.2. Multisensor and multiresolution classification 21.1.3. Previouswork 51.2. Methodology 91.2.1. Overview of the proposed approaches 91.2.2. Hierarchical model associated with the first proposed method 101.2.3. Hierarchical model associated with the second proposed method 131.2.4. Multisensor hierarchical MPM inference 141.2.5. Probability density estimation through finite mixtures 171.3. Examples of experimental results 191.3.1. Results of thefirstmethod 191.3.2. Results of the secondmethod 221.4. Conclusion 261.5. Acknowledgments 261.6. References 27Chapter 2. Pixel-based Classification Techniques for Satellite Image Time Series 33Charlotte PELLETIER and Silvia VALERO2.1. Introduction 332.2. Basic concepts in supervised remote sensing classification 352.2.1. Preparing data before it is fed into classification algorithms 352.2.2. Key considerations when training supervised classifiers 392.2.3. Performance evaluation of supervised classifiers 412.3. Traditional classification algorithms 452.3.1. Support vector machines 452.3.2. Random forests 512.3.3. k-nearest neighbor 562.4. Classification strategies based on temporal feature representations 592.4.1. Phenology-based classification approaches 602.4.2. Dictionary-based classification approaches 612.4.3. Shapelet-based classification approaches 622.5. Deep learningapproaches 632.5.1. Introduction to deep learning 642.5.2. Convolutionalneuralnetworks 682.5.3. Recurrentneuralnetworks 712.6. References 75Chapter 3. Semantic Analysis of Satellite Image Time Series 85Corneliu Octavian DUMITRU and Mihai DATCU3.1. Introduction 853.1.1.TypicalSITSexamples 893.1.2. Irregular acquisitions 903.1.3.The chapter structure 963.2.Why are semantics neededin SITS? 963.3.Similaritymetrics 973.4. Feature methods 983.5. Classification methods 983.5.1. Active learning 993.5.2. Relevance feedback 1003.5.3. Compression-based pattern recognition 1003.5.4. LatentDirichlet allocation 1013.6. Conclusion 1023.7. Acknowledgments 1053.8. References 105Chapter 4. Optical Satellite Image Time Series Analysis for Environment Applications: From Classical Methods to Deep Learning and Beyond 109Matthieu MOLINIER, Jukka MIETTINEN, Dino IENCO, Shi QIU and Zhe ZHU4.1. Introduction 1094.2. Annual time series 1114.2.1. Overview of annual time series methods 1114.2.2. Examples of annual times series analysis applications for environmentalmonitoring 1124.2.3. Towardsdense time series analysis 1164.3. Dense time series analysis using all available data 1174.3.1. Making dense time series consistent 1184.3.2. Change detection methods 1214.3.3. Summaryand futuredevelopments 1254.4. Deep learning-based time series analysis approaches 1264.4.1. Recurrent Neural Network (RNN) for Satellite Image TimeSeries 1294.4.2. Convolutional Neural Networks (CNN) for Satellite Image TimeSeries 1314.4.3. Hybrid models: Convolutional Recurrent Neural Network (ConvRNN) models for Satellite Image Time Series 1344.4.4. Synthesis and future developments 1364.5. Beyond satellite image time series and deep learning: convergence between time series and video approaches 1364.5.1. Increased image acquisition frequency: from time series to spacebornetime-lapse and videos 1374.5.2. Deep learning and computer vision as technology enablers 1384.5.3. Future steps 1394.6. References 140Chapter 5. A Review on Multi-temporal Earthquake Damage Assessment Using Satellite Images 155Gül¸sen TA¸SKIN, Esra ERTEN and Enes OØguzhan ALATA¸S5.1. Introduction 1555.1.1. Research methodology and statistics 1595.2. Satellite-based earthquake damage assessment 1655.3. Pre-processing of satellite images before damage assessment 1675.4. Multi-source image analysis 1685.5. Contextual feature mining for damage assessment 1695.5.1. Textural features 1705.5.2. Filter-based methods 1735.6. Multi-temporal image analysis for damage assessment 1755.6.1. Use of machine learning in damage assessment problem 1765.6.2. Rapid earthquake damage assessment 1805.7. Understanding damage following an earthquake using satellite-based SAR 1815.7.1. SAR fundamental parameters and acquisition vector 1855.7.2. Coherent methods for damage assessment 1885.7.3. Incoherent methods for damage assessment 1925.7.4. Post-earthquake-only SAR data-based damage assessment 1955.7.5. Combination of coherent and incoherent methods for damage assessment 1965.7.6. Summary 1985.8. Use of auxiliary data sources 2005.9. Damage grades 2005.10. Conclusionand discussion 2035.11. References 205Chapter 6. Multiclass Multilabel Change of State Transfer Learning from Image Time Series 223Abdourrahmane M. ATTO, Héla HADHRI, Flavien VERNIER and Emmanuel TROUVÉ6.1. Introduction 2236.2. Coarse- to fine-grained change of state dataset 2256.3. Deep transfer learning models for change of state classification 2326.3.1. Deep learningmodel library 2326.3.2. Graphstructures for theCNNlibrary 2346.3.3. Dimensionalities of the learnables for the CNN library 2366.4. Change of state analysis 2376.4.1. Transfer learning adaptations for the change of state classification issues 2386.4.2. Experimental results 2396.5. Conclusion 2436.6. Acknowledgments 2446.7. References 244List of Authors 247Index 249Summary of Volume 1 253

O autorze

Abdourrahmane M. Atto is Associate Professor at the University Savoie Mont Blanc, France. His research interests include mathematical methods and models for artificial intelligence and image time series.Francesca Bovolo is the Head of the Remote Sensing for Digital Earth Unit, Fondazione Bruno Kessler, Italy. Her research interests include remote sensing image time series analysis, content-based time series retrieval and radar sounders.Lorenzo Bruzzone is Professor of Telecommunications and the Founder and Director of the Remote Sensing Laboratory at the University of Trento, Italy. His research interests include remote sensing, machine learning and pattern recognition.