An Improved Attributed Scattering Model Optimized by Incremental Sparse Bayesian Learning

Abstract

In this paper, we propose an improved attributed scattering model to mathematically unify the scattering models of several canonical primitives. These primitives include not only point- and line-segment-scatterers, such as trihedral, cylinder, dihedral, and rectangular plane, but also arc scatterers, such as sphere and top-hat. The estimation of the model parameters can be posed as an ill-posed linear inverse problem. To overcome the ill-posedness, we employ the incremental sparse Bayesian learning method to realize the sparsity-driven continuous parameter estimation. Inverse scattering experiments demonstrate that the proposed methodology not only provides desirable sparse representations of the target scattering response but is also able to capture richer geometrical information than the existing models.