Automatic Region-Based Shape Discrimination Of Ground Penetrating Radar Signatures

Automatic processing and interpretation of ground-penetrating radar data has been the obsession of radar practitioners and geophysicists for several years, offering the hope of overcoming the bottleneck that often limits the practicality and cost-effectiveness of this tool for rapid site investigation. This paper presents a technique for rapid and consistent interpretation of GPR data achieved by training an artificial neural network to discriminate between data-segments originating from different types of targets and other undesired sources of reflections such as clutter. In addition to that, the network is trained to distinguish between hyperbolic-shaped reflections (such as pipes and tanks) and non-hyperbolic-shaped ones (such as soil disturbances). This is achieved by computing a number of discriminating regional shape properties from B-scan segments. The neural classifier was combined with some image enhancement, edge detection and thinning operations to produce a system that is capable of returning 3-dimensional images highlighting regions of extended targets (such as storage tanks and soil disturbances) and pinpointing the location of localised targets such as mines and pipes. This classifier was applied to a variety of GPR data sets gathered from a number of sites. The obtained results were in close agreement with those obtained by a trained operator manually, but in a fraction of the time.