Superpixel-Level Target Discrimination for High-Resolution SAR Images in Complex Scenes

Abstract

Traditional synthetic aperture radar (SAR) target discrimination methods are implemented at the chip-level, which may have good discrimination performance in simple scenes but may lose the effectiveness in complex scenes. To improve the discrimination performance in complex scenes, this paper proposes a superpixel-level target discrimination method directly in high-resolution SAR images. The proposed discrimination method mainly contains three stages. First, based on the superpixel-level target detection results, we describe each superpixel via the multilevel and multidomain feature descriptor, which can reflect the differences between targets and clutter comprehensively. Second, we employ the support vector machine as the discriminator to obtain the discriminated target superpixels. Finally, we cluster the discriminated target superpixels and extract the target chips from the original SAR image based on the clustering results. The experimental results based on the miniSAR real SAR data show that the proposed discrimination method has about 25% higher F1-score than the traditional discrimination methods.