Multi-temporal PolSAR Image Classification Using F-SAE-CNN

Abstract

Crop classification using polarimetric SAR data is one of the most important applications in Polarimetric Synthetic Aperture Radar (PolSAR) imagery. Obviously, for crop classification, multi-temporal PolSAR data can provide more information than single-temporal PolSAR data, but the processing method of the matching image data is relatively backward. Aiming at the high-dimensional data composed of multi-temporal PolSAR, this paper proposes a method to integrate the stacked auto-encoder network and convolutional neural network, making full use of the dimension reduction advantages of the stacked auto-encoder network and the superior classification performance of the convolutional neural network. By constructing a fusion network, the multi-temporal PolSAR images can be processed once, the classification accuracy can be improved, and the processing steps can be simplified. The experimental results show that, compared with the traditional Stacked Auto-encoder and Convolutional Neural Network (SAE-CNN) classification method, the multitemporal PolSAR image classification method based on Fusion of Stacked Auto-encoder and Convolutional Neural Network (F-SAE-CNN) proposed in this paper has the highest classification accuracy, which effectively combines the advantages of the self-encoding network and the CNN network, and provides a new idea for PolSAR image classification work.