Model-based Automated Target ID and Pose Estimation from SAR Data

Automatic Target Recognition (ATR) and the determination of target intent from airborne Synthetic Aperture Radar (SAR) depends on highly accurate pose determination. In this proposal prior knowledge, in the form of 3D target models, is applied to achieve ‘super resolution’; namely a resolution beyond the physical constraints of radar data acquisition and processing. With GPS positioning and high bandwidth SAR, the accurate locating of a few (<10) major scatterers in correct geometrical relationship should allow for post-processing of SAR images to achieve a five degree orientation accuracy. This proposal will test the conditions for which a sphere scatterer model is sufficiently accurate to simulate the contribution of individual scatterers to the radar signature for 3D pattern matching. Starting with an established SSI-developed, hardbody rendering code, QUick Image Display (QUID), we extract the facet normals to determine sphere radii assigned to model vertices. Because the radar signature from a sphere is known exactly, QUID model range maps can be employed to determine an approximate radar signature in real time for pattern matching. During Phase I the SSI proposed target pose determination code, Augmented QUID Synthetic Aperture Radar (QUASAR), will be tested on Army SAR data and 3D CAD models.