An automatic feature extraction technique for GPR data processing in electromagnetic inverse scattering problems

Abstract

This work presents an automatic feature extraction technique for the analysis of GPR data. Despite the rapidly increasing demand for such technology, the interpretation of radargrams is still a specific task which only expert users can accomplish. As a radar probing system, in fact, the GPR is able to measure the backscattered EM waves from buried targets and the travel time of these signals, information which cannot be directly exploited, but that needs to be be used in inverse scattering problems for the evaluation of the dielectric properties and spatial positions of subsurface targets. Starting from a practical example of GPR employment — the reconstruction of pavement's vertical cross section — the evaluation of the two above-mentioned features is carried out through an energy-based method, which ensures high performances and robustness to signal interference.