Feature extraction of sar target in clutter based on peak region segmentation and regularized orthogonal matching pursuit

Feature extraction in clutter is a challenging problem in SAR target recognition because of the difficulty in distinguishing the target signature from the background. In this paper, a new feature extracting algorithm based on automated peak region segmentation (PRS) and regularized orthogonal matching pursuit (ROMP) techniques is presented and called PRS-ROMP. It combines the processes in both signal domain and image domain. First, the proposed method uses PRS and parametric model (PM) to obtain the positions and atoms of strong scattering centers of target. Then we acquire the positions and atoms of weak scattering centers by the sparse reconstruction algorithm and PM for residual region. By using all atoms of strong and weak scattering centers we get the final amplitude estimation by LS. Experimental results of electromagnetic calculations data in clutter validate the proposed target feature extraction method.