Localization and backscattering density estimation from GPR data with neural network

Abstract

An adaptive linear neuron network is employed for reversing the location and back scattering density of objects from ground penetrating radar data. The processing avoids the disadvantage of unknown electromagnetic velocity in a medium for the specific rebar detecting application. Based on the common-offset reflection GPR survey model, the network was derived by reconstructing and compressing the reflected signal matrix. The location and scattering density of the targets under investigation are extracted by fitting the output of the network to the measured data. Finally, experiments with high-resolution configurations confirmed the reliability of the proposed method, and further developments are discussed.