Terrain Segmentation in PolSAR Images via Statistical Learning and Uncertainty Perception

Recently, deep learning network is introduced to terrain segmentation application in polarimetric synthetic aperture radar (PolSAR) images and has achieved remarkable performance. However, interclass and intraclass varieties caused by nonlinear characteristic and intrinsic randomness introduced by speckle are still a challenge for segmentation methods. In this letter, a segmentation framework based on nonlinear statistical learning is introduced to bridge the gap between terrain segmentation in PolSAR and optical images. First, a learnable nonlinear statistical description module (LSDM) is proposed to represent the nonlinear characteristic introduced by coherent speckle, which supplements the lack of discriminative linear feature in terrain blocks. Second, a conditional diffusion pipeline is introduced to model latent distributions and perceive intrinsic randomness that causes intraclass variety. Finally, the proposed framework adopts the statistical features learned from nonlinear statistical description module as a condition to align with the process of distribution modeling in the conditional diffusion segmentation pipeline (CDSP). Extensive experiments are conducted on two authoritative PolSAR terrain segmentation datasets, which present competitive performance.