Probabilistic maps for buried pipes location based on GPR images

Abstract

In this work, statistical methodologies are used to generate automatic probabilistic maps of pipe layouts in water supply systems (WSS) based on street surveys using ground penetrating radar (GPR). Probabilistic maps are based on the analysis of GPR images along with a multi-agent generation of pseudo-random walks, and a process to discard areas with less probability of pipe existence. This is an iterative procedure that we have integrated into a system that produces GPR sampling walks, and eventually set up a reliable location map of buried pipes. As a result the survey time is optimized and the amount of data needed to conduct records of the components of WSS is minimized.