[HTT18-01] Automatic estimation of the position of buried pipes based on the spatial distribution of semblance
Keywords:ground penetrating radar, 3D geophysics, buried pipe
A hyperbolic curve appears on the B-scan image drawn by scanning to the transverse direction against a buried pipe. Since symmetric and continuous patterns are characteristic of the buried pipes and rare in the natural underground, the reflected wave from the buried pipe is easy to identify. The engineering geophysicists have so far mapped the position of buried pipes using this property. Here, if we regard the observed hyperbolic curve as the time curve traveled from a single point reflector, we can obtain the propagation velocity so that the semblance calculated along the travel time curve is maximized. We call the propagation velocity the apparent velocity. We conducted highly dense GPR survey at our test sites and analyzed the buried pipes. The survey lines were placed not only the longitudinal direction but the transversal direction. Using the data, we searched the semblance as the values maximize while changing the apparent velocity and extracted the local maximum values from the spatial semblance distribution. Although the obtained point clouds of the semblance contained not only reflections from buried pipes but also the other scatters in the soils, the points by buried pipes tended to show higher semblance. The results of simulating the travel time curve in case of the survey line located orthogonal to the direction of cylindrical pipe in subsurface showed almost equivalent to a point reflection, however when closing the crossing angle between the survey line and the cylindrical pipe down to 45 degree, the apparent velocity was up to about 1.5 times of actual velocity. We tried to filter for passing only the point clouds of the buried pipes. Using the constraints of the simulated apparent velocity range, we searched the cloud which the number of the counts aligned on the line connected between a pair of high semblance points is maximum. And we removed the isolated points. Consequently, we could automatically extract the point clouds of buried pipes (Fig. 1).
In next tasks, we have a plan to cluster the filtered point clouds to each buried pipe group and insert the objects correspond to specification of the pipe. We are developing the automatic method of the clustering. We think that the estimated depth errors of the reconstructed 3-D model will be improve because the propagation velocities will be more accurate by considering the crossing angle between the survey line and the buried pipe.