Hough transform based automatic pipe detection for array GPR: Algorithm development and on-site tests

GPR is one of the most reliable instrument to obtain information on underground assets. Since its diffusion, many efforts have been made in order to overcome one the main limitations of this technology: the ldquohuman factorrdquo in data interpretation. Actually, operators have to look and interpret a huge number of radargrams. This requires an enormous effort, about 2 days are usually needed for analyzing radar data gathered from a 3000 m
²
wide site Some useful tools like tomography has been developed, but their quality and reliability is related to the migration algorithm which is heavily depending on knowledge of the propagation velocity of EM wave in the soil. Hyperbolas automatic detection algorithm reduces human factor in radar data interpretation and aims to strongly lower the amount of time needed for data analysis. The algorithm works on each radargram, recognizes the presence of target by looking for the hyperbolic patterns and marks them. Then the operator can correlate those results from different radar maps and extract the position of the underground assets. The algorithm also gives an accurate evaluation of the propagation velocity of EM waves in the soil, so this estimate can be used by migration algorithm in order to obtain an improved tomography. Using both Hyperbolas Automatic Detection and tomography, time needed to analyze a 3000 m
²
wide area can be reduced up to less than 4 hours.