Rectangle outline extraction of vehicles in SAR images

The outline of a vehicle in Synthetic Aperture Radar (SAR) images can be modeled as a rectangle. The rectangle outline of vehicles is beneficial in suppressing ground clutter with less loss of target signatures, and brings benefit to vehicle classification using SAR images. Moreover, the rectangle outline provides a suitable context for representing subparts of vehicles, and is therefore, profitable for classifying vehicles under Extend Operation Conditions (EOC). The critical problems that handicap the accurate rectangle outline extraction are fake objects due to speckle noise and clutter, missing objects resulting from weak backscattering, and possible articulations such as tank gun barrels. In this paper, prior information about shadow boundary and vehicles´ length-width ratio is introduced to attack these problems, and a heuristic method is proposed to extract the rectangle outline of vehicles. Experiment results on MSTAR dataset show the validity of the proposed method.