Determination of Parameters of Subsurface Layers Using GPR Spectral Inversion Method

A step-by-step spectral inversion algorithm is proposed in this paper to estimate the thickness, depth, permittivity, and conductivity of layered media from ground-penetrating radar (GPR) signals. Due to the distinct GPR spectral responses to media parameters, different attributes of the spectra are used to estimate, separately, one or several parameters within each step. Such estimated parameters are then used as starting values in a nonlinear optimization problem for the inversion of all parameters simultaneously from the whole GPR spectral data. The cost function of the optimization problem is defined as the difference between the modeled generalized reflection coefficient spectrum of subsurface layers and the actual one. The optimization problem is solved by using a modified stochastic hill-climbing (SHC) algorithm. The method is applied on synthetic data of a wedge model and real data from two highway GPR detections. The results show that the step-by-step spectral inversion method can reduce the ambiguity of the inversion and improve its accuracy and efficiency.