Change Detection Algorithm Based on U-NET Convolutional Neural Network for Multitemporal Wavelength-Resolution SAR Images

This paper presents a change detection algorithm (CDA) for multitemporal wavelength-resolution synthetic aperture radar (SAR) images based on U-NET convolutional neural network (CNN) architecture. The algorithm uses non-linear filtering to extract textural information from amplitude SAR images. Since we want to detect changes in multitemporal acquisitions, relevant textures (entropy and variance) can be obtained from the histograms of sum and difference SAR images. The original images and the obtained textures are given as input of the proposed algorithm. As a result, the proposed CDA has a similar detection performance compared to other methods, but the number of false alarms is significantly reduced.