A Multiscale Optimization Technique for Large-Scale Subsurface Profiling

Abstract

An efficient multiresolution inverse scattering approach is presented for profiling high-contrast buried targets in large investigation domains (IDs). The proposed technique is based on an iterative multiscale approach (IMSA) that starts with coarse meshes and successively zooms in and marches toward the detected target location. Hence, by tightening the ID, the resolution could be enhanced without any increase in the number of meshes of the search domain. Here, the global evolutionary programming (EP) optimization algorithm is used in each step of IMSA, to guarantee the success of inversion process. In particular, an efficient variation in EP with Cauchy mutation is implemented for enhanced convergence. Moreover, a hybrid combination of multifrequencies (MFs) and frequency hopping (FH) schemes is combined with the proposed technique to better deal with the problem’s nonlinearity and also simultaneously make a consistent trade-off between mesh length and frequencies. The proposed FH-MF-IMSA is evaluated by implementing it to various large-scale subsurface problems. Furthermore, the superior performance of this technique is shown by comparing its results with the standard IMSA, as well as with other well-known global optimization techniques.