Automatic Subsurface Map Generation based on GPR Data Processing

In the domain of subsurface scanning, an abundance of GPR data are typically acquired by exploiting technologically advanced Ground Penetrating Radar (GPR) devices. However, the processing and representation of these data in a concise manner that can be easily interpreted by non-experts is still an active research topic. To this end, the paper at hand introduces a new method for automatic registration of sequences of GPR data into a common reference frame. It starts from an elementary GPR scanning session, relied on the existence of a GPR antenna array in a structured device, and proposes a sequential registration approach for multiple elementary scans, by exploiting a similarity measure, that utilizes the multilevel discrete wavelet transform decomposition. The data are processed in the 3D-space to achieve convergence during merging the scans, formulating thus a 3D map of the subsurface. In this map, data of selected depth range are top-down projected to formulate C-Scans in a parametric manner that can be easily interpreted by non-experts. The method has been evaluated in three different datasets and exhibited promising results in the task of automated GPR scans registration and corresponding large-scale subsurface map generation.