A Geometrical-Based Simulator for Target Recognition in High-Resolution SAR Images

Target recognition in high-resolution synthetic aperture radar (SAR) images is a challenging task. In this letter, a novel geometrical-based SAR image simulator is proposed to assist target recognition. Specifically, in addition to the using of Lamberitian-specular mixed model for single-bounce simulation, we propose a dihedral corner model for double-bounce simulation, which allows directly retrieving structure information from scattering patterns. In addition, a dihedral tracing technique is taken instead of ray tracing for double-bounce detection to reduce computation time. Also, a new primitive-by-primitive visualization approach which is well combined with our scattering model is proposed to obtain high efficiency. The efficiency of the simulator is demonstrated by the comparison between the simulated results and MiniSAR images and by the application in vehicle recognition.