Adaptive Processing of SAR Data for ATR

摘要

In this presentation, we outline a framework for Automatic Target Recognition (ATR) in Synthetic Aperture Radar (SAR) that is based on an adaptive processing of the digitally-spotlighted phase history data of individual targets. In the conventional SARATR algorithms, the magnitude only or the complex SAR image of a test target chip is analyzed and compared with a set of reference target chips to determine the test target type or class. Depending on the relative coordinates of a test target in the imaging scene and the flight path of the radar-carrying aircraft, the SAR image (chip) of the test target possesses a spatial warping with respect to the reference target chips that should be blindly incorporated and/or compensated in the ATR algorithm. Furthermore, the formed SAR images exhibit certain (slant plane) parametric variations and erroneous shifts that depend on the type of the SAR imaging algorithm that is used. For the success of a SAR-ATR algorithm, the spatial warping and slant plane parameters of the SAR imaging algorithm should be available to the user. In practice, these parameters of the formed SAR image are not exactly known to the SAR-ATR user.