Elimination of leakage and ground-bounce effects in ground-penetrating radar data

We address the problem of removing specular ground surface reflections and leakage/cross-talk from downward looking stepped frequency ground-penetrating radar (GPR) data. A new model for the ground-bounce and the leakage/cross-talk is introduced. An algorithm that jointly estimates these effects from collected data is presented. The algorithm has the sound foundation of a non-linear least squares (LS) fit to the presented model. The minimization is performed in a cyclic manner where one step is a linear LS minimization and the other step is a non-linear LS minimization where the optimum can efficiently be found using, e.g., the chirp-transform algorithm. The results after applying the algorithm to measured GPR data, collected at a U.S. army test range, are also shown.