SAR Image Classification Using Greedy Hierarchical Learning With Unsupervised Stacked CAEs

Zhensheng Sun,Juan Li,Peng Liu,Weijia Cao,Tao Yu,Xingfa Gu

DOI: https://doi.org/10.1109/tgrs.2020.3023192

IF: 8.2

2021-07-01

IEEE Transactions on Geoscience and Remote Sensing

Abstract:Synthetic aperture radar (SAR) can provide stable data source for earth observation due to its advantages of all day and night, all-weather, and strong penetration. SAR image classification as a fundamental procedure has been proved its great value in plenty of remote sensing applications. Conventional classification algorithms mainly rely on hand-designed features, which are susceptible to widespread coherent speckle noise and geometric distortion in high-resolution SAR images. Inspired by the recent impressive success in data mining and deep learning, a greedy hierarchical convolutional neural network (GHCNN) is developed. It aims at obtaining optimized feature representation, relieving the effect of speckle noise, and promoting the local pattern recognition of geometric distortion in single-polarized SAR image classification. First, a series of convolutional autoencoders (CAEs) is trained in the greedy layer-wise unsupervised strategy. This step provides an unbiased regularizer and a priori distribution derived from large volumes of unlabeled SAR patches. Then, to optimize multiple parameter subspaces globally, several CAEs are coupled together to form a deeper hierarchical structure in a stacked and unsupervised fashion. Afterward, a convolutional network with identical topology inherits the pretrained weights. After supervised finetuning, it realizes class prediction. Synchronously, t-distributed stochastic neighbor embedding (t-SNE) algorithm is applied to monitor the efficiency of feature representation during the training period. Experimental results demonstrate that the proposed method has competitive advantages over involved contrast methods.

imaging science & photographic technology,remote sensing,engineering, electrical & electronic,geochemistry & geophysics

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve The paper primarily aims to address several key issues in synthetic aperture radar (SAR) image classification: 1. **Challenges in Feature Extraction**: Traditional classification algorithms rely on manually designed features, which are susceptible to the prevalent speckle noise and geometric distortions in high-resolution SAR images. 2. **Optimizing Feature Representation**: A Greedy Hierarchical Convolutional Neural Network (GHCNN) is proposed to obtain optimized feature representations, mitigate the impact of speckle noise, and enhance local pattern recognition under geometric distortions. 3. **Dataset Limitations**: The amount of SAR image data is limited, while training deep models typically requires a large amount of labeled data. Therefore, the paper proposes a semi-supervised training mechanism to make full use of the limited labeled data. Through the above methods, the paper aims to improve the accuracy and robustness of single-polarization SAR image classification, demonstrating competitiveness in various application scenarios.

SAR Image Classification Using Greedy Hierarchical Learning With Unsupervised Stacked CAEs

A Novel Group Squeeze Excitation Sparsely Connected Convolutional Networks for SAR Target Classification

SAR Image Classification Using Fully Connected Conditional Random Fields Combined with Deep Learning and Superpixel Boundary Constraint

Simulation-assisted SAR Target Classification Based on Unsupervised Domain Adaptation and Model Interpretability Analysis

A CA-Based Weighted Clustering Adversarial Network for Unsupervised Domain Adaptation PolSAR Image Classification

Adaptive Hierarchical Multinomial Latent Model with Hybrid Kernel Function for SAR Image Semantic Segmentation

Deep Convolutional Generative Adversarial Network With Autoencoder for Semisupervised SAR Image Classification

Multi-Scale Convolutional Neural Network for SAR Image Semantic Segmentation

SAR Image Classification Via Hierarchical Sparse Representation and Multisize Patch Features

Classification of Large-Scale High-Resolution SAR Images with Deep Transfer Learning

PolSAR Image Classification Based on Multi-Scale Stacked Sparse Autoencoder.

SAR image classification using deep features and neighborhood heterogeneity confidence refined spatial constraints

Classification for SAR Scene Matching Areas Based on Convolutional Neural Networks.

High-Resolution SAR Image Classification Using Multi-Scale Deep Feature Fusion and Covariance Pooling Manifold Network

Robust Unsupervised Small Area Change Detection from SAR Imagery Using Deep Learning

HENC: Hierarchical Embedding Network With Center Calibration for Few-Shot Fine-Grained SAR Target Classification

SAR Target Classification Based on Integration of ASC Parts Model and Deep Learning Algorithm

SAR-CDSS: A Semi-Supervised Cross-Domain Object Detection from Optical to SAR Domain

A Feature Fusion-Net Using Deep Spatial Context Encoder and Nonstationary Joint Statistical Model for High-Resolution SAR Image Classification

SAR Image Segmentation Based on Hierarchical Visual Semantic and Adaptive Neighborhood Multinomial Latent Model

Synthetic aperture radar remote sensing scene classification based on fuzzy co-occurrence networks