Discontinuity Identification in Rock Masses Using a Region-growing-based Method: A Case Study Using the RockBench Dataset

Abstract

Rock mass discontinuities exert a fundamental control on the mechanical behavior and stability of rock engineering structures. With the increasing availability of high-resolution point clouds acquired by LiDAR and photogrammetry, automated identification of rock mass discontinuities has become an important topic in rock engineering and geomatics [4,5]. However, the performance of existing automatic methods is often sensitive to data quality, and systematic validation using standardized benchmark datasets remains limited. In this study, a region-growing-based method is implemented for the identification of rock mass discontinuities from three-dimensional point cloud data. Local normal vectors and curvature are estimated using principal component analysis, seed points are selected based on surface smoothness, and regions are progressively grown under constraints of normal vector consistency and spatial continuity. The method is evaluated using a publicly available benchmark dataset from the RockBench repository [1]. The results demonstrate that the proposed method can effectively delineate major discontinuity planes and extract their geometric parameters, including orientation and spatial extent. Statistical analysis reveals several dominant discontinuity sets that are consistent with previously reported geological characteristics. A parameter sensitivity analysis further illustrates the robustness and limitations of the method. This study highlights the applicability of region-growing-based discontinuity identification for rock mass characterization using high-resolution point cloud data.