Change Alignment-Based Image Transformation for Unsupervised Heterogeneous Change Detection

Change detection (CD) with heterogeneous images is currently attracting extensive attention in remote sensing. In order to make comparable, the image transformation methods transform one into domain of another image, which can simultaneously obtain a forward difference map (FDM) and backward (BDM). However, previous only fuse FDM BDM post-processing stage, cannot fundamentally improve performance CD. this paper, change alignment-based (CACD) framework for unsupervised CD proposed deeply utilize complementary information process, enhances effect transformation, thus improving performance. To reduce dependence network on labeled samples, we propose graph structure-based strategy generating prior masks guide network, influence changing regions an way. More importantly, based fact that are representing same event, perform alignment during enhance enable effectively indicate real region. Comparative experiments conducted six state-of-the-art five datasets, showing CACD achieves best average overall accuracy (OA) 95.9% different datasets at least 6.8% improvement kappa coefficient.