Preface
Contents
End-to-End ATR Leveraging Deep Learning
1 Overview
2 Introduction
3 Front-End Algorithms
3.1 Target Detection
3.1.1 CFAR Prescreener
3.1.2 Target vs. Clutter Discriminator
3.2 Target Orientation Estimation
4 Classification Algorithm
4.1 Algorithm Training
4.2 Validation Experiment
4.2.1 The Learned Network
4.2.2 Comparison to the Literature
4.2.3 Comparison to Template Matching
5 Conclusion
References
Change Detection in SAR Images Using Deep Learning Methods
1 Introduction
2 State of the Art
2.1 Convolutional Autoencoders
2.2 Change Detection
2.3 CD Methods Using SAR Images
2.4 Unsupervised DL CD Method Using SAR Images
3 Unsupervised Change Detection in SAR Images Using Deep Learning Multi-scale Features
3.1 Unsupervised Change Detection Based on Convolutional Autoencoder Feature Extraction
3.1.1 Unsupervised CAE Training
3.1.2 Feature Extraction
3.1.3 Change Detection
4 Experimental Design and Results
4.1 Description of Datasets
4.2 Design of the Experiments
4.3 Experiment 1: Analysis of the Performance Varying the Number of Layers
4.4 Experiment 2: Analysis of the Performance Varying the Window Size of the Standard-Deviation-Based Reliability Approach
4.5 Experiments 3 and 4: Brumadinho Dataset
4.6 Experiments 3 and 4: L'Aquila Dataset
5 Conclusions
References
Homography Augmented Momentum Contrastive Learning for SAR Image Retrieval
1 Introduction
2 Related Works
3 Methodology
3.1 Contrastive Learning
3.2 Homography Transformation
3.3 Training
4 Numerical Experiments
4.1 SAR Image Data
4.2 Experiment Settings
4.3 Experiments Results
5 Conclusions
References
Synthetic Aperture Radar Image Based Navigation Using Siamese Neural Networks
1 Introduction
2 Model Formulation
2.1 Contrastive Loss and Triplet Loss
2.2 Dataset Generation and Retrieval
3 Numerical Experiments and Results
3.1 Data and Architecture Details
3.2 Computational Results and Discussion
3.3 Case Study
4 Conclusions
References
A Comparison of Deep Neural Network Architectures in Aircraft Detection from SAR Imagery
1 Introduction
2 Datasets
3 Methods
3.1 SSD
3.2 Faster R-CNN
3.3 EfficientDet
3.4 Efficient Weighted Feature Fusion and Attention Network (EWFAN)
3.5 YOLOv5
3.6 Efficient Bidirectional Path Aggregation Attention Network (EBPA2N)
4 Results and Discussion
4.1 Aircraft Detection in Airport I
4.2 Aircraft Detection in Airport II
4.3 Aircraft Detection in Airport III
4.4 Evaluation and Analysis of Detection Performance
5 Conclusion and Future Works
References
Machine Learning Methods for SAR Interference Mitigation
1 Introduction
1.1 Congested Electromagnetic Environment
1.2 Adverse Impacts of Interference to SAR Systems
1.2.1 Data Collection Process
1.2.2 Image Formation Process
1.2.3 Image Interpretation Process
1.3 Notations
1.4 Summary of the Remainder of the Chapter
2 Interference Mitigation Strategies for SAR System
2.1 Interference Detection
2.2 Overall Introduction of Interference Suppression Methods
2.2.1 Nonparametric Methods
2.2.2 Parametric Methods
2.2.3 Semi-parametric Methods
3 Machine Learning Methods for Interference Suppression
3.1 General Signal Model of Typical Interferences for SAR systems
3.1.1 Narrowband Interference
3.2 Wideband Interference
3.3 Interference Mitigation Schemes Based on Machine Learning
3.3.1 NBI Mitigation
3.3.2 WBI Mitigation
3.3.3 Simultaneous Mitigation of Complicated NBI and WBI
4 Future Trend of Interference Mitigation via Deep Learning and Cognitive Scheme
References
Classification of SAR Images Using Compact Convolutional Neural Networks
1 Introduction
2 Background and Related Work
2.1 SAR Data Processing
2.2 PolSAR Information Extraction
2.2.1 Coherent Target Decompositions
2.2.2 Incoherent Target Decompositions
2.3 Prior Work
2.3.1 Classification of Partially Polarized SAR Data
2.3.2 PolSAR Data Classification
3 Methodology
3.1 Adaptive CNN Implementation
3.2 Back-Propagation for Adaptive CNNs
4 Experimental Results
4.1 Benchmark SAR Data
4.1.1 Po Delta, COSMO-SkyMed, X-band (PDelta_X)
4.1.2 Dresden, TerraSAR-X, X-band (Dresden_X)
4.1.3 San Francisco Bay, AIRSAR, L-band (SFBay_L)
4.1.4 San Francisco Bay, RADARSAT-2, C-band (SFBay_C)
4.1.5 Flevoland, AIRSAR, L-band (Flevo_L)
4.1.6 Flevoland, RADARSAT-2, C-band (Flevo_C)
4.2 Experimental Setup
4.3 Results
4.3.1 Results on PDelta_X
4.3.2 Results on Dresden_X
4.3.3 Results on SFBay_L
4.3.4 Results on SFBay_C
4.3.5 Results on Flevo_L
4.3.6 Results on Flevo_C
4.3.7 Deep Versus Compact CNNs
4.3.8 Sensitivity Analysis on Hyper-Parameters
5 Conclusions and Future Work
References
Multi-Frequency Polarimetric SAR Data Analysis for Crop Type Classification Using Random Forest
1 Introduction
2 Study Area and Dataset
3 Methodology
3.1 Random Forest
3.2 Parameter Importance Evaluation
3.3 Partial Probability Plot
3.4 Processing Steps for Parameter Selection and Classification Using RF
4 Results and Discussion
4.1 Separation Among Long-Stem (LS) Crops
4.2 Separation Among Short-Stem Broad-Leaf (SSBL) Crops
4.3 Separation Between SSBL and LS Crops
4.4 Analyzing the Mixing Among Crop Classes
5 Summary
References
Automatic Determination of Different Soil Types via Several Machine Learning Algorithms Employing Radarsat-2 SAR Image Polarization Coefficients
1 Introduction
2 Materials
2.1 Pilot Area
2.2 Measurement of Local Soil Type Samples
2.3 Acquisition of Radarsat-2 SAR Image Data
2.4 Radarsat-2 SAR Image Preprocessing
3 Methods
3.1 Feature Extraction
3.2 K-Nearest Neighbor (K-NN) Algorithm
3.3 Extreme Learning Machine (ELM) Algorithm
3.4 Naive Bayes (NB) Algorithm
3.5 Performance Metrics
4 Results and Discussion
4.1 Determination of Soil Types via Polarimetric SAR Coefficients and K-NN
4.2 Determination of Soil Types via Polarimetric SAR Coefficients and ELM
4.3 Determination of Soil Types via Polarimetric SAR Coefficients and NB
5 Conclusion
References
Ocean and Coastal Area Information Retrieval Using SARPolarimetry
List of Abbreviations
1 Radar Polarimetry
1.1 Polarimetric Scattering Descriptors
1.2 PolSAR Imaging Modes
1.3 Sea Surface Polarimetric Scattering
1.4 Experimental Showcases
2 SAR Polarimetry for Sea Oil Spill Observation
2.1 Overview
2.2 PolSAR for Marine Oil Spill Observation
2.2.1 Feature Extraction to Monitor Sea Oil Spills
2.2.2 CP SAR Architectures
2.2.3 Challenges and Research Trend
2.3 Experimental Showcase
2.3.1 SAR Polarimetry for Sea Oil Spill Observation: Conventional Classifiers
2.3.2 SAR Polarimetry for Sea Oil Spill Observation: Convolutional Neural Network Classifiers
3 SAR Polarimetry for Shoreline Monitoring
3.1 State-of-the-Art
3.2 Methodology
3.3 Experimental Showcase
4 Conclusions
References