A data-driven multidimensional signal-noise decomposition approach for GPR data processing

We demonstrate the possibility of applying a data-driven nonlinear filtering scheme in processing ground penetrating radar (GPR) data. The algorithm is based on the recently developed multidimensional ensemble empirical mode decomposition (MDEEMD) method which provides a frame of developing a variety of approaches in data analysis. The GPR data processing is very challenging due to the large data volume, special format, and geometrical sensitive attributes which are very easily affected by various noises. Approaches which work in other fields of data processing may not be equally applicable to GPR data. Therefore, the MDEEMD has to be modified to fit the special needs in the GPR data processing. In this study, we first give a brief review of the MDEEMD, and then provide the detailed procedure of implementing a 2D GPR filter by exploiting the modified MDEEMD. A complete synthetic model study shows the details of algorithm implementation. To assess the performance of the proposed approach, models of various signal to noise (. S/. N) ratios are discussed, and the results of conventional filtering method are also provided for comparison. Two real GPR field examples and onsite excavations indicate that the proposed approach is feasible for practical use.