Estimation of Discontinuities from Point Cloud Based on Variable-Box Segmentation Method

Abstract

For estimating discontinuities of a rock mass from point cloud (LiDAR) data, an algorithm called DiAnahas been used. It obtains a fracture plane from the valid point cloud inside a cubic bounding box. In DiAna, to extract the valid point cloud from the cubic bounding box, DiAna has to determine the threshold value to remove noises. It seems that this manual operation is hard work for long tunnels. To improve this manual operation, we developed Variable-Box Segmentation (VBS) method for LiDAR data recorded from tunnel. VBS has three processes, i.e., first segmentation, second segmentation, and combining. During first segmentation, the point cloud is segmented into large bounding boxes and noise is removed. During second segmentation, each box is divided into nine sub-boxies. Planes are estimated from point cloud data inside each sub-box. During combining, sub-boxies containing similar planes are joined. VBS was examined using LiDAR data including three sets of major discontinuities. Estimation results from VBS was compared with the reference planes decided from geological sketch. Results showed that similarity between reference planes and planes determined by VBS algorithm seems to be enough to find discontinuities from fractured planes.