SAR Target Segmentation Based On Shape Prior Constraint

With the development of synthetic aperture radar(SAR) technology, SAR has been an important branch of remote sensing field. Target segmentation is the base of SAR image interpretation and quite meaningful for the SAR image research. However, the segmentation is challenging due to the speckle noise and target’s heterogeneity. The introduction of shape priors to the segmentation has been proved very effective in optical image. This paper introduces shape priors to the level set method to segment SAR target, and the main works are as following1) Analyze the geometric features and region features of SAR targets, then propose the definition of SAR target shape, and finally introduce the shape prior to the level set method.2) Propose an improved level set method based on shape prior. The traditional level set function(LSF) is declined to be distorted, so a novel function is introduced to maintain the regularity of LSF, and therefore the re-initialization of LSF is avoided. Segmentations of the ship and building validate the algorithm and the robustness to the noise.3) The nonlinear transformation of the target’s shape exists due to the angle of the imaging sensor. So the model combines the image information and shape priors is proposed, and kernel principle component analysis(KPCA) is used to extract the nonlinear features to construct the shape prior constraint energy. Parzen windows can be used to estimate the target area and background area and then the image information energy is given. Experiments of MSTAR vehicles validate the proposed algorithm.