A Multiresolution Hybrid Technique for Large-Scale Multiple Buried Targets Profiling

Abstract

An efficient hybrid inverse scattering technique for dealing with large-scale half-space GPR scenarios, is proposed. First, an initialization step using a fast and simple Diffraction Tomography (DT) method has been utilized to roughly detect the objects' borders and reduce the size of the investigation domain (ID). Then, an iterative multi-scale approach (IMSA) is applied that starts with coarse meshes, and then successively zooms inside the regions where the objects are located. In each step, the inversion procedure is performed by the stochastic optimization method of Covariance Matrix Adaption Evolutionary Strategy (CMA-ES) which is a robust technique for nonlinear problems. Furthermore, multi frequency data are used in each inversion process, to more effectively alleviate the problem's ill-posedness. The presented results show the reliability and efficiency of this technique in dealing with multiple buried targets. Moreover, comparison with a standard IMSA approach, reveals that the proposed scheme significantly outperforms it.