Performance Analysis for Synthetic Aperture

In recent years, synthetic aperture radars (SARs) have been used to detect man-made targets and to distinguish them from naturally occurring background. The purpose of this thesis is to assess target classification performance of a SAR-based automatic target recognition (ATR) system starting from a foundation of rigorous, physics-based signal models developed from electromagnetic scattering theory without incurring the restrictive assumptions made in previous work. Targets consist of a repertoire of geometrically-simple reflectors, and the performance discrepancy of a conventional full-resolution processor with respect to a multi/adaptive-resolution processor is discussed. The analytical computation of target classification performance is generally demanding due to the dense correlation between the likelihood values of different targets. Thus most of our attention is devoted to obtaining upper and lower bounds on probability of correct classification (PCC). Three types of target conditions are investigated: (1) targets consisting of a known constellation of reflector components at known absolute locations; (2) targets consisting of a constellation of known reflector components which are located at random positions within some limited, prescribed uncertainty regions; and (3) targets consisting of a constellation of known reflector components with known centroid location but unknown rotation about that centroid. For settings 1 and 2, we obtain a lower bound on PCC from the performance of a recognition processor that makes component-wise reflector decisions, and we obtain an upper bound on PCC by assuming that the returns from a target's reflector components have known relative phases. Computer simulations show that the lower bound is very close to the exact result. For setting 3, we use the Laplace approximation to obtain an approximation to PCC that is valid at high signal-to-noise ratios. Thesis Supervisor: Professor Jeffrey H. Shapiro Title: Julius A. Stratton Professor of Electrical Engineering