Progressive compressive sensing for exploiting frequency-diversity in GPR imaging
Résumé
The microwave imaging of buried targets from wide-band ground penetrating radar (GPR) signals is addressed. By considering a contrast source inversion (CSI) formulation of the inverse scattering equations and taking advantage of both the intrinsic frequency diversity of GPR data and the sparseness of the unknown buried scatterers within the subsurface domain of investigation, a multi-task Bayesian compressive sensing (MT-BCS) approach is integrated within a frequency hopping (FH) inversion scheme. Towards this end, an innovative “constrained” relevance vector machine (C-RVM) solver is developed to effectively exploit the information, progressively acquired at each frequency step, on the unknown scattering scenario. Representative numerical benchmarks and preliminary experimental results are presented to assess the effectiveness and the potentialities of the proposed subsurface imaging method (namely, the FH-MT-BCS technique).