Optical-ISAR (Inverse Synthetic Aperture Radar) image conversion method and device based on hierarchical domain prior

The invention relates to an optical-ISAR image conversion method and device based on hierarchical domain prior. The method comprises the following steps: by proposing a hierarchical domain prior-based optical-radar satellite ISAR image generation model, namely a cross-modal conversion network, on the basis of a cyclic consistency theory, the network combines scattering physical domain prior of an ISAR image and classification domain prior oriented by a classification task to construct brand new scattering consistency loss and classification consistency loss, so that the scattering consistency loss and the classification consistency loss of the ISAR image are improved. According to the method, the loss function and the iteration framework of the model are optimized, the model learns the mapping relation between the optical image domain and the ISAR image domain by minimizing the network loss function based on the established satellite target optical image data training sample with the CycleGAN as the backbone