Method and system for determining main orientation of feature blocks in visible and SAR images based on generative adversarial networks

Abstract

The present invention relates to a method and system for determining the main orientation of feature blocks in visible and SAR images based on generative adversarial networks. The method includes: acquiring multiple pairs of heterogeneous image groups; accurately locating the exact corresponding coordinate points in each pair of segmented visible and SAR image groups respectively based on HOPC and FSC methods; clipping the segmented visible and SAR image groups with the corresponding coordinate points as the centers respectively, and randomly rotating the clipped SAR images by an angle of ; inputting the clipped visible and SAR image groups into an initial GAN network for weight training to obtain the common features of heterogeneous images.