Scholarlink

<p>Rock mass discontinuities fundamentally govern the mechanical behavior and stability of engineered rock structures. However, automated characterization of these features from 3D point clouds remains challenging due to three critical limitations: (1) the reliance on manual specification of discontinuity set numbers, (2) inadequate integration of spatial continuity and orientation similarity, and (3) limited capability for multi-scale structural characterization. This study proposes an integrated framework that combines density peak clustering algorithm (DPCA), adaptive spatial–orientation -weighted clustering, and hierarchical region growth refinement to address these challenges in a unified and fully data-driven manner. The framework enables automatic determination of discontinuity set numbers through density distribution analysis of normal vectors, thereby eliminating subjective parameter selection. Validation using benchmark Rockbench datasets demonstrates that the proposed method achieves orientation deviations within 2° for dip direction and 3° for dip angle relative to reference measurements, remaining well within acceptable geological error bounds for engineering applications. Compared with baseline methods, computational efficiency is improved by approximately 72%, accompanied by a 28% enhancement in boundary delineation accuracy. The proposed framework provides an effective and scalable solution for automated rock mass characterization, with direct implications for slope stability assessment, tunnel design, and geotechnical risk evaluation.</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

Automatic Recognition of Rock Mass Discontinuities from 3D Point Clouds: An Integrated Density Peak Clustering and Multi-scale Refinement Framework

  • 期刊论文

  • 发表日期：

    2026年5月25日

查看全文
  • Fei Xue,
  • Jinyu Yan,
  • Jinhua Qian,
  • Xiangzhen Wang,
  • Xiaoqing Wang,
  • Tianzuo Wang,
  • Zhiyi Zhang,
  • Junhao Mi

摘要

Rock mass discontinuities fundamentally govern the mechanical behavior and stability of engineered rock structures. However, automated characterization of these features from 3D point clouds remains challenging due to three critical limitations: (1) the reliance on manual specification of discontinuity set numbers, (2) inadequate integration of spatial continuity and orientation similarity, and (3) limited capability for multi-scale structural characterization. This study proposes an integrated framework that combines density peak clustering algorithm (DPCA), adaptive spatial–orientation -weighted clustering, and hierarchical region growth refinement to address these challenges in a unified and fully data-driven manner. The framework enables automatic determination of discontinuity set numbers through density distribution analysis of normal vectors, thereby eliminating subjective parameter selection. Validation using benchmark Rockbench datasets demonstrates that the proposed method achieves orientation deviations within 2° for dip direction and 3° for dip angle relative to reference measurements, remaining well within acceptable geological error bounds for engineering applications. Compared with baseline methods, computational efficiency is improved by approximately 72%, accompanied by a 28% enhancement in boundary delineation accuracy. The proposed framework provides an effective and scalable solution for automated rock mass characterization, with direct implications for slope stability assessment, tunnel design, and geotechnical risk evaluation.