SAR Image Change Detection Based On URNet Network.

In order to improve the accuracy of synthetic aperture radar (SAR) image change detection based on deep convolution neural network. In this paper, a SAR image change detection method based on U-Rank Network (URNet) is proposed. Firstly, the convolution unit of U-shaped network is used to extract the features of the input image, and then the size of the input features is scaled by down-sampling(DS) part and up-sampling(US) part. Secondly, inspired by squeeze and excitation(SE) network, rank operation is added on the basis of SE module, and the relationship between input feature information channels is further used to enhance channel information. Then, the output features of each branch are restored to the same size by resampling unit. Finally, the features of each branch are fused together, and the final result graph is obtained through the whole connection layer. The experimental results of three groups of different types of SAR show that the method proposed in this paper is superior to other compared methods.