Shadow-based Moving Target Detection For Video SAR

Abstract

Direct energy return of moving target, represented as a fast-shifting bright spot in Video Synthetic Aperture Radar (ViSAR) imaging, may leave shadow in its true location. This phenomenon can be utilized for slow MTD (Moving Target Detection). Whereas the variation caused by the bright spot and the serious speckle noise will degrade the detection effect and bring in excessive false alarms. This paper proposes a novel shadow-based ViSAR MTD method including both preprocessing and moving target detection procedure. In preprocessing step, an efficient inter-frame registration algorithm based on SURF (Speeded-Up Robust Feature) is realized to avoid the long interval misregistration problem. The V-BM3D (Video Block-Matching 3D) algorithm is introduced to smooth the background speckle noise. Then in the following detection procedure, an algorithm fusing background difference and symmetric difference is proposed and implemented on ViSAR image sequence to extract more area of moving target shadow. False alarms caused by displaced bright spots are suppressed via gray value discrimination. Experiments on a real ViSAR data are carried out and the results validate the remarkable detection performance.