Full-waveform Inversion of Crosshole Radar Based on Matlab Distributed Computer Engine (MDCE)

Crosshole ground penetrating radar (GPR) tomography plays an important role in geoscientific and engineering investigations. Traditional tomography (i.e., fitst-arrival times and maximum first-cycle amplitudes) based on ray theory can't provide high resolution images because only a fraction of the information contained in the data is used in the inversion. Here, we realize a full-waveform inversion (FWI) which can exploit all information present in the observed data and the method can provide sub-wavelength image. The inversion technology is the steepest descent method based on a 2-D finite-difference time-domain solution of Maxwell's equations. We build a complex model which has three layers and two sub-wavelength pipes embedded in the middle layer to test the effect of the method. Results of synthetic data show that the full waveform tomography can accurately depict the interface of layers and the size of sub-wavelength bodies. Finally, we successfully apply FWI to the field data, and the results of FWI are compared to the results from applying ray-based.