Automatic Rock Mass Discontinuities Detection Based on Binocular 3-D Reconstruction

Measurement of rock mass discontinuities was the key for rock mass 3D network simulation, seepage analysis, and stability analysis. In this paper, a non-contact rock mass discontinuities detection method based on the single camera binocular three-dimensional reconstruction theory was presented. This method provided an effective way for grouping rock mass discontinuities and calculation the discontinuity orientations. The main flow of the method was: (1) obtaining three dimensional point cloud data based on the binocular three-dimensional reconstruction theory by single camera, (2) de-noising and re-sampling on point cloud data before triangulation to reduce holes in the reconstructed triangular mesh, (3)grouping rock mass discontinuities automatically based on an improved K-means algorithm which adopted density and clustering validity indexes, (4) segmenting the discontinuities in the same group based on the angle and the adjacent relationship between two triangular facets, and (5) fitting the segmented point cloud using the Random Sample Consensus (RANSAC) algorithm and calculating its orientation. This method was applied to the grouping and measurement of discontinuities of a rock mass slop. The results showed that the grouping and measurement of discontinuities were reliable and of high accuracy that could meet the engineering requirements.