关键词: 结构面单面, 优势组结构面, 改进区域生长算法, S-K-means聚类, 智能识别

Abstract: The discontinuities of natural rock mass possess specific mechanical properties that can define the vulnerable parts of the rock mass. These properties play a decisive role in the structure, strength and stability of various rock engineering projects, such as tunnel support, classification of surrounding rock mass, and slope reinforcement. Therefore, it is of crucial importance to recognize the single discontinuity and the dominant groups with relatively developed conditions. In proposed method, the automatic identification steps of dominant groups of discontinuities are divided into three parts, point clouds normal vectors calculation, single discontinuity segmentation and dominant groups clustering: 1) Calculate normal vectors based on Robust Randomized Hough Transform; 2) An improved region-growing algorithm is proposed to segment a number of discontinuities. In terms of seed points selection and region-growing conditions, curvature, planarity and roughness are considered, and dynamic outlier detection is added. In addition, the relationship between the thresholds and the number of discontinuities is used to qualitatively judge the extreme segmentation situation, and the optimal threshold range is screened out; 3) Finally, S-K-means clustering algorithm is proposed to recognize dominant groups clustering. The accuracy of the algorithm is verified by a rock slope. The results demonstrate that the inclination angle error ranges from 0.7° to 2.5°, and the average inclination angle error is 1.8° and 1.7° respectively. This method shifts from directly clustering point clouds to identify dominant groups to first segmenting several single discontinuity before clustering. This refinement enhances the robustness and accuracy of discontinuities clustering computation, increases computational speed, and maintains applicability across various discontinuities datasets. Consequently, this provides a more precise and rapid method for the intelligent identification of discontinuities.

Key words: single discontinuity, dominant groups of discontinuities, improved region-growing algorithm, S-K-means cluster, intelligent identification