Surface Parameter Estimation by Inverse Modeling of Ground Penetrating Radar

Abstract

Ground Penetrating Radar is a high resolution electromagnetic technique to analyse the soil sub-surface. This requires accurate forward modeling of the GPR soil sub-surface system. Depending upon the application and the environment in which GPR is to be employed several forward analytical or numerical modeling methods are proposed by the researchers. Some of the analytical methods include standard reflection coefficient modeling, full waveform inversion modeling, transmission line modeling etc. Inverse modeling of GPR, which is getting back the electromagnetic parameters of targets based on the electromagnetic responses of the soil sub-surface, requires objective function to be defined and optimized. The objective function is defined as an error function which is minimum for the actual soil parameters. The accuracy and efficiency of the GPR system depends on the accuracy of the forward modeling and accuracy and efficiency of the inverse modeling. In this thesis work the common reflection coefficient method is explored for forward modeling and a spectral domain inversion technique is implemented for a layered ground surface. A three layered synthetic model of ground media is defined. The objective function is defined as an error function between the actual reflection coefficient and the modeled reflection coefficient. The variation of the objective function for different soil parameters like permittivity, height, conductivity is studied. Finally soil parameters are extracted with good accuracy by using global search techniques like Pattern Search and GA combined with Nelder-Mead Simplex method. It is also observed that the step by step detection of few parameters of first and second layer based on the properties of the reflection coefficient spectrum enhances the process of extracting complete electrical profile of the ground sub-surface.