Automatic extraction of rock discontinuity orientations from 3D point clouds via an adaptive clustering and surface fitting approach

Abstract

The orientation of rock discontinuities is a critical parameter for evaluating the stability and safety of rock engineering structures. With the continuous advancement of remote surveying techniques, analysis of exposed rock surfaces based on 3D point cloud data has emerged as a mainstream approach, owing to its high data fidelity and rich geometric information. However, efficiently and accurately extracting geometric parameters of rock discontinuities from point clouds remains a significant challenge. To address this issue, this study proposes an efficient method for the automatic extraction of geometric features of rock discontinuities from 3D point clouds. First, a downsampling and smoothing preprocessing strategy is employed to significantly enhance computational efficiency while preserving essential geometric features. Concurrently, a local geometric adjustment normal estimation algorithm is introduced to generate high-precision normals while retaining sharp structural features. An improved Unsupervised K-means (UKM) clustering algorithm is subsequently proposed to planar segmentation of the point cloud, enabling the automatic identification and classification of discontinuity orientations. Finally, an enhanced Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm is adopted to achieve precise planar segmentation, followed by Random Sample Consensus (RANSAC) method to ensure accurate extraction of discontinuity orientations. Experiments conducted on two real-world datasets demonstrate that the proposed method outperforms four widely used approaches in terms of both accuracy and computational efficiency, providing a novel and effective solution for the automated extraction of structural surface information in rock engineering applications.