Predictive Rotation Fusion: A Physical Model-Based Fusion Method for Full-Polarimetric GPR Data

Since the applications of the full-polarimetric ground penetrating radar (FP-GPR) technique have evolved from the detection of isolated targets to the interpretation of regional geologic structures, the study of the FP-GPR data fusion imaging method has become an urgent issue. Current data fusion methods for FP-GPR are all mathematical algorithm-based, aiming to obtain optimal weights for stacking. However, these stacking methods will produce wrong results if there are phase oppositions between different channels of FP-GPR data. Therefore, this article proposes a physical model-based fusion method called predictive rotation fusion (PRF). The new method aims to pursue special angles between the antenna and the surveying line for every scattering point in the subsurface, at which the amplitudes of acquired ground penetrating radar (GPR) data will be maximum. Besides, a new predictive filter is established and embedded into the PRF method to suppress the distortions and noises in the fusion results. The effectiveness of the proposed method is verified through laboratory experiments and its superiorities in fusion and denoising are also discussed by comparing it with a currently best weight-based fusion method. Finally, two real experiments are introduced to show the abilities of the PRF method in urban underground space imaging and field loess structure imaging.