Fusion of multiple time‐domain GPR datasets of different center frequencies

Abstract

ABSTRACTGround‐penetrating radar is widely used in non‐destructive underground inspection because of its precision and high resolution, especially at shallow layers. However, single‐antenna ground‐penetrating radar data may lead to erroneous evaluations. To avoid this problem, we propose a fusion method for ground‐penetrating radar data acquired at different center frequencies. The datasets are first preprocessed using zero calibration and denoizing, and subsequently processed further using a novel algorithm for spatial (horizontal and vertical) calibration that produces results in the same spatial coordinate system. We carried out a comparative fusion experiment using Fourier, wavelet, S and principal‐component transforms and using data acquired at center frequencies 100, 200 and 400 MHz. Our results show that the proposed fusion algorithm can effectively exploit the complementarity and redundancy of information at different center frequencies, with the wavelet‐transform‐based fusion showing the best performance on image quality according to different metrics. Moreover, the proposed method enhanced the information entropy and the average gradient, whereas the root‐mean‐square error of the fusion data remained below 24%.