PolSAR Image Classification Based on Deep Convolutional Neural Network

For introducing the advantages of feature learning and multilayer network in the interpretation of Polarimetric synthetic aperture radar (PolSAR) image, a classification algorithm based on deep convolutional neural network is proposed, and is used for PolSAR image classification. Firstly, a special convolutional neural network (CNN) for PolSAR image is constructed, secondly, a large number of PolSAR training data is introduced into the multilayer network, and the CNN network is trained adequately, thirdly, the test data is imported into the multilayer network, and the classification of polarimetric SAR images is realized. Experimental results on the first batch of PolSAR data show that the proposed algorithm can achieve higher classification accuracy rate.