Physics-Driven CSI for High-Dimensional SAR

Realizing the promise of 3D from a single-aperture, single-pass requires a new generation of processing and exploitation algorithms conceived and optimized for 3D SAR from sparse-aperture measurements. We propose an alternative, physics-driven constrained sparse inverse algorithm that provides consistent, interpretable reconstructions and precludes the user from having to select model order or regularization parameters. Our approach combines the physical interpretability of attributed scatterers with the “plug-in” convenience of non-parametric regularized inverse algorithms. We apply recently developed theory in sparse function recovery via convex l1 optimization and constrain the solution to be a convex combination of attributed scatterer and similar functions. Since the resulting optimization problem is convex, the solution is guaranteed to be unique, avoiding the need to select user-driven parameters.