A Physics-Based Imaging and Feature Extraction Technique for Bistatic Radar Imaging

We present a feature extraction technique for bistatic radar imaging based on exact nonlinear anisotropic scattering representations to overcome the isotropic assumption in traditional radar imaging. We accomplish this by incorporating angle-dependent basis sets within an imaging framework. We model each unknown scatterer in a scene as the linear superposition of known bistatic pose and scale invariant bases. We define a novel anisotropic sparse imaging algorithm that replaces the isotropic assumption with our anisotropic basis set introducing an angular dependency within the imaging frame-work. We exploit a multi-resolution approach to estimate both the scale and location of the scatterer for scale invariance. Our algorithm exploits the proximal gradient descent algorithm to find the optimal mapping from the basis set to the spatial do-main. We present results for various scatterers. We validate the assumption of sparsity by comparing against its l
2
-norm regularized counterpart.