Back to search

University of British Columbia

Efficient automatic extraction of discontinuities from rock mass 3D point cloud data using unsupervised machine learning and RANSAC

Abstract

dc:description

Geometric characterization of discontinuity planes in rock masses is an essential practice in rock engineering. Analysis of 3D point cloud data of exposed rock surfaces acquired by remote sensing techniques is an increasingly adopted approach for discontinuity plane identification and orientation measurement. This research proposes a computationally efficient workflow to extract planar discontinuities from a point cloud automatically. The proposed workflow includes voxel-grid downsampling as a preprocessing step to smooth the point cloud and reduce the number of points for processing. A covariance matrix is then formed, and singular value decomposition is applied to estimate normal vectors. Next, primary clustering is used to assign the normal vectors into a set of clusters. Then secondary clustering is performed to separate individual discontinuity planes in each set. Finally, plane fitting using RANSAC to extract the orientation, location, and equation of each discontinuity plane. Analysis of multiple point cloud data sets of a rock face was used to demonstrate the proposed workflow capability to extract discontinuities accurately and calculate their orientations in a computationally efficient procedure. It is demonstrated that the proposed automated procedure is superior in terms of computational run time, the accuracy of orientation results, and the efficiency for processing large point cloud data sets. Thorough comparisons with published results and analysis of performance metrics for the segmented planes are also performed. The results obtained from the proposed automated procedure for a well-studied point cloud data set show an average discontinuity orientation discrepancy of around 2° versus the ground truth results.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy - PhD
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Civil Engineering
Grantor dc:publisher
University of British Columbia
Year dc:date
2022

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Daghigh, Hamid
Contributors dc:contributor
  • Tannant, Dwayne

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language
eng

Identifiers

dc:identifier.*
Handle dc:identifier
http://hdl.handle.net/2429/81760

Chain of custody

source
Harvested from
University of British Columbia
Base URL
circle.library.ubc.ca/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Daghigh, Hamid. Efficient automatic extraction of discontinuities from rock mass 3D point cloud data using unsupervised machine learning and RANSAC. doctoral thesis, University of British Columbia, 2022. http://hdl.handle.net/2429/81760