Automatic Fracture Orientation Extraction from SfM Point Clouds

摘要

Geology seeks to understand the history of the Earth and its surface processes through charac- terisation of surface formations and rock units. Chief among the geologists’ tools are rock unit orientation measurements, such as Strike, Dip and Dip Direction. These allow an understanding of both surface and sub-structure on both the local and macro scale.\nAlthough the way these techniques can be used to characterise geology are well understood, the need to collect these measurements by hand adds time and expense to the work of the geologist, precludes spontaneity in field work, and coverage is limited to where the geologist can physically reach.\nIn robotics and computer vision, multi-view geometry techniques such as Structure from Motion (SfM) allows reconstructions of objects and scenes using multiple camera views. SfM-based techniques provide advantages over Lidar-type techniques, in areas such as cost and flexibility of use in more varied environmental conditions, while sacrificing extreme levels of fidelity. Regardless of this, camera based techniques such as SfM, have developed to the point where accuracy is possible in the decimetre range.\nHere is presented a system to automate the measurement of Strike, Dip and Dip Direction using multi-view geometry from video. Rather than deriving measurements using a method applied to the images, such as the Hough Transform, this method takes measurements directly from the software generated point cloud.\nPoint cloud noise is mitigated using a Mahalanobis distance implementation. Significant structure is characterised using a k-nearest neighbour region growing algorithm, and final surface orientations are quantified using the plane, and normal direction cosines.