Two fast buried pipe detection schemes in Ground Penetrating Radar images

In this paper two fast detection schemes for Ground Penetrating Radar data analysis are presented. Both processing chains are able to detect small underground targets (usually pipes) by recognizing the hyperbolic signatures that they produce and both are based on pattern recognition methods. In particular, a neural network and a template matching approach were implemented, allowing a fast post-acquisition data interpretation. The detection schemes are also able to cluster the found signatures into different classes with respect to their shape (which, in turn, provides more information on the targets). Experimental results validating the approach on two test sites in Turin and Nantes are presented.