Discontinuous surface extraction method based on 3D point cloud

In underground engineering, precise analysis of structural discontinuities is critical for understanding the rock fracture mechanisms subjected to shear and tensile loading. This study presents an automatic method for identifying structural planes based on 3D point cloud data of sandstone. The methodology integrates K-nearest neighbor (KNN) search and random sample consensus (RANSAC) algorithms to compute normal vectors, followed by mean shift clustering for preliminary grouping and Euclidean clustering for discontinuity orientation. Key parameters (dip angle, trend, and area) of dominant discontinuities are systematically extracted and quantified. In order to verify the accuracy of the method, two engineering cases (regular hexahedron and rock slope) are selected for analysis. The results show that this method has high consistency in dip angle and trend extraction, which can automatically extract small-scale structural planes in complex rock strata and accurately calculate their area which is superior to traditional methods in terms of accuracy and robustness. The parameter selection (bandwidth = 0.4, distance threshold = 0.3, and screening threshold = 200) balances computational efficiency and precision, reducing over-segmentation while preserving critical structural details. The research results can provide theoretical guidance for engineering fields such as slope stability evaluation and crack propagation simulation.