Exploitation of Frequency Diversity in GPR Imaging through an Innovative Constrained-BCS Method

This work deals with the retrieval of the electromagnetic (EM) properties of buried domains starting from wide-band ground penetrating radar (GPR) measurements. The GPR inverse scattering (IS) problem is formulated within the contrast source framework and solved through an innovative multi-task Bayesian compressive sensing (MT-BCS)-based approach. More precisely, the developed methodology integrates a customized MT-BCS solver inside a frequency-hopping (FH) inversion strategy. Thanks to a novel "constrained" relevance vector machine (C-RVM) algorithm, progressively acquired information about the solution through the cascaded processing of different frequency components can be effectively exploited. Some numerical results are shown in order to assess the effectiveness of the proposed FH-MT-BCS method, as well as to compare it against a single-task (FH-STBCS) implementation not exploiting the correlation among different illuminations.