Abstract:

The effectiveness of cross-hole radar tomography depends mainly on the quality of the extracted first arrival-time. Digital image segmentation method, based on the projection onto convex sets (POCS) technique, is used to extract the first arrival-time by segmenting the color image of the energy ratio, and had previously been applied in refracted seismic first arrival-time extraction. We first applied digital image segmentation method into cross-hole radar travel-time tomography to reconstruct the velocity field using an iteratively linearized inversion approach. During the inversion, LSQR algorithm was employed to solve the system of linear equations, Jacobian matrix was constructed by the curved ray tracing technique, and the traveltime was calculated using Multistencils Fast Marching Method (MSFM). We employed a synthetic data set and a field data set to test the effectiveness of the digital image segmentation method in traveltime tomography. For comparison, a traditional energy ratio method is also used for first arrival-time extraction. The result reflected that the tomography based on digital image segmentation method is more accurate with smaller residuals, and can provide more effective judgment for the underground velocity field.

Key words: cross-hole radar, first arrival-time extraction, travel-time tomography, digital image segmentation, projection onto convex sets