Advanced Geostructural Survey Methods Applied to Mock Mass Characterization

摘要

The location and orientation of rock discontinuities, which are traditionally obtained from geological surveys with obvious drawbacks (safety, rock face accessibility, etc.), may also be derived from a detailed and accurate photogram-metric or laser scanning survey. Selecting from the point cloud determined on the rock face a set of points distributed on a particular discontinuity, location, dip, and dip direction can be computed from the least-squares estimate of the plane interpolating the set of points. Likewise, the normal vector to the surface may be computed from an interpolation or approximation of the surface by appropriate functions. To become a real alternative (both in terms of productivity as well as accuracy) to a traditional survey, interactive or automated software tools are necessary, to allow the efficient selection of the point sets on the discontinuities or the interpretation of the normal vector pattern. After introducing the two best technologies available today for data acquisition and their performance, the paper presents an approach, based on the random sample consensus (RANSAC) procedure, to the segmentation of the point cloud into subsets, each made of points measured on a discontinuity plane of the rock face. For each subset, the plane's equations coefficients are first determined by robust estimation and then refined by least-squares estimation after outlier removal. The segmentation algorithm has been implemented in RockScan, a software tool developed to facilitate the interaction with the point cloud in the identification of the discontinuities; rather than using the three-dimensional (3D) data, selection of regions of interest is performed on oriented images of the rock face. Finally, application of RockScan to four different lest sites is discussed and results presented. The sites differ in size (from tens to hundreds of meters), rock surface characteristics, and the technology used to produce the point cloud (in three cases photogrammetry, in the fourth laser scanning), giving the opportunity to test the methodology in different contexts. In the first and in the fourth site an extensive traditional survey has been performed, providing reference data to validate the RockScan results.