GNGTS 2019 - Atti del 38° Convegno Nazionale

702 GNGTS 2019 S essione 3.2 highlighted from the background, thus making possible quantitative considerations about their dimensional parameters and geometry. Semi-automatic picking . For some pipelines, semi-automatic interpretation was very quick and effective: for these cases it was enough to pick only few control points. On the contrary, in other cases it was necessary to pick closer control points, having a mean distance not exceeding 4 m. Picked results were carefully evaluated on both maps and within the entire 3D volume (Fig. 2). We accurately imaged even small depth variations of pipelines (e.g. yellow and orange pipes in the center of Fig. 2) as well as their spatial continuity. The geometry at the crossing points was further reconstructed and the semi-automatic procedure was able to identify the pipes even where they intersect or they are very close. All the picked features are georeferenced and can therefore be exported and easily plotted on different base maps (Fig. 3). Conclusion. We analyzed the applicability of 14 volume attributes, calculated on a dense 3D GPR dataset in order to implementing a procedure for semi-automatic mapping of underground utilities. Some of the tested attributes demonstrate their effectiveness to highlight position, direction and depth variations of pipes. Results show that with dense and high-quality data just few control points are required to obtain an accurate picking, in which the user defined thresholds are quite robust. Only few outliers are picked but they can be easily identified and Fig. 1 - Envelope Time slice (Two way travel time equal to 25.415 ns) with main underground utilities marked by yellow arrows. Red circle shows an area characterized by materials different from the surroundings. Blue line indicates the direction of acquisition.