Scholarlink

<p>Accurate identification of rock mass discontinuities is a prerequisite for reliable geotechnical stability analysis and numerical modeling. In this study, a highly efficient and fully automated workflow was developed for detecting discontinuities from 3D point clouds by integrating the Neighborhood Reweighted Local Centroid (NRLC) feature extraction scheme with a two-stage DBSCAN clustering framework. Initially, boundary, concave, and convex feature points are identified and filtered to mitigate noise interference. Discontinuity sets are then extracted in the normal vector space, followed by a merging strategy to address the ambiguity of normal vector orientation. Subsequently, spatial clustering is applied to segment individual discontinuities, and for complex orientations, the two-stage clustering refinement algorithm (ATSCRA) is introduced to recover structural planes that may have been misclassified as noise. Case studies on three benchmark datasets indicate that the proposed method achieves high accuracy in orientation estimation, particularly for rock point clouds with complex orientations, where the average error in dip and dip direction for the majority of planes is below 0.4° compared to the benchmark values. These results confirm the robustness, accuracy, and scalability of the proposed method, providing a reliable tool for large-scale characterization of discontinuities in complex geological environments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

NRLC-Enhanced Two-Stage DBSCAN Clustering for Rock Discontinuity Identification by 3D Point Clouds

  • 期刊论文

  • 发表日期：

    2025年11月25日

查看全文
  • Yuanzhen Xu,
  • Shuqi Ma,
  • Xiangchen Yao,
  • Zhaoyuan Zhang,
  • Jiazheng Chen

摘要

Accurate identification of rock mass discontinuities is a prerequisite for reliable geotechnical stability analysis and numerical modeling. In this study, a highly efficient and fully automated workflow was developed for detecting discontinuities from 3D point clouds by integrating the Neighborhood Reweighted Local Centroid (NRLC) feature extraction scheme with a two-stage DBSCAN clustering framework. Initially, boundary, concave, and convex feature points are identified and filtered to mitigate noise interference. Discontinuity sets are then extracted in the normal vector space, followed by a merging strategy to address the ambiguity of normal vector orientation. Subsequently, spatial clustering is applied to segment individual discontinuities, and for complex orientations, the two-stage clustering refinement algorithm (ATSCRA) is introduced to recover structural planes that may have been misclassified as noise. Case studies on three benchmark datasets indicate that the proposed method achieves high accuracy in orientation estimation, particularly for rock point clouds with complex orientations, where the average error in dip and dip direction for the majority of planes is below 0.4° compared to the benchmark values. These results confirm the robustness, accuracy, and scalability of the proposed method, providing a reliable tool for large-scale characterization of discontinuities in complex geological environments.