Principal Investigator Alan Willsky
magery formed from wide-angle synthetic aperture radar (SAR) measurements has fine cross-range resolution in principle. However, conventional SAR image formation techniques assume isotropic scattering, which is not valid with wide-angle apertures. Also, the spatial location of scattering centers may migrate as a function of viewing angle across the aperture. The problem of jointly forming images and characterizing anisotropy as well as characterizing scattering center migration in wide-angle SAR is considered in the research. The approach not only compensates for anisotropy and migration in the image formation process, but gives much more information, useful for scene interpretation, than a simple image would.
A method based on a sparse representation of anisotropic scattering with an overcomplete dictionary composed of atoms with varying levels of angular persistence is presented. Solved as an inverse problem, the result is a complex-valued, aspect-dependent response for each scatterer in a scene. The non-parametric approach jointly considers all scatterers within one system of equations. The choice of the overcomplete dictionary incorporates prior knowledge of anisotropy, but the method is flexible enough to admit solutions that may not match a family of parametric functions. Sparsity is enforced through regularization based on the l_p quasi-norm, p<1, leading to a non-convex minimization problem. A quasi-Newton method is applied to the problem and a novel graph-structured algorithm is developed and also applied. Results are demonstrated on synthetic examples and realistic examples with XPatch data, including the backhoe public release dataset.