A Novel Two-Step Registration Method for Multi-Aspect SAR Images

Synthetic aperture radar (SAR) images of metal targets are highly sensitive to the observation angle, which is unbeneficial for target recognition. To solve this problem, a novel two-step image registration method is proposed in this paper. Firstly, multiple-aspects SAR images are projected into a unified coordinate system to complete the coarse image registration by using GPS information of the flight. Secondly, a fine SAR image registration method named GI-SIFT (Geographic information-scale invariant feature transform) is presented, which combines geographic information and scale-invariant feature transform algorithm. After registration, multi-aspects SAR images are fused to obtain multi-aspect scattering features of the target, thus improve the performance of targets recognition. Furthermore, the experiment results using real data demonstrate the effectiveness and benefits of the proposed method.