Use of Compressive Sensing in Efficient Multiscale Stochastic-based Inverse Profiling of Multilayered Subsurface Targets

Abstract

A multiresolution electromagnetic inverse scattering algorithm is proposed for reconstruction of multiple objects buried in a multilayered media, using limited amount of data. For tackling this highly nonlinear and ill-posed problem, it is required to combine different inverse profiling modalities. The proposed algorithm starts with using the total variation minimization (TVM) method to locate the objects and roughly estimate their borders. Then, an iterative multi-scale approach (IMSA) is implemented to confine the investigation domain (ID) to the detected objects regions, step by step. Using this approach, the resolution can be enhanced without increasing the number of unknowns. As the inverse solver, the robust stochastic optimization technique of Covariance Matrix Adaption Evolution Strategy (CMA-ES) is utilized in each step. The presented numerical results indicate the efficiency of the proposed algorithm in quantitative profiling of multiple objects, using limited amount of data.