Noise-robust target recognition of SAR images based on attribute scattering center matching

A noise-robust target recognition method of synthetic aperture radar (SAR) images is proposed based on attributed scattering center (ASC) matching. The ASCs are extracted based on sparse representation, which keep stable under possible noise corruption. In the classification stage, the test and corresponding template ASC sets are matched using the Hungarian algorithm to build a one-to-one correspondence. Then, a similarity measure is defined to measure the similarity of the test and template ASC sets, which also considers the noise corruption. Based on the calculated similarities, the target label is recognized as the class achieving the highest similarity. By using the noise-robust features and designing the noise-robust classification scheme, the proposed method can achieve better robustness to possible noise corruption. The Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset is used for experimental evaluation of the proposed method under both the standard operating condition (SOC) and different levels of noise corruption.