Feature extraction of a generic SAR target using an improved data model

Abstract

Here feature extraction of a generic SAR target is studied. We apply a new data model to target feature extraction and improve SAR image quality. The data model we construct contains (1) envelope, (2) location of the target. The envelope shape can be controlled by four parameters, and location information is indicated by frequency pair in both range and cross-range. These parameters would be estimated by nonlinear least squares (NLS) and 1-D Cramer-Rao Bounds (CRB) for these parameters have also been analyzed. Numerical examples show this method can achieve CRB at high SNR and its computational complexity is acceptable.