Abstract: To rapidly and efficiently obtain the orientation information of rock mass discontinuities, and to address the challenges of parameter selection difficulties and poor cross-scenario applicability in current mainstream point cloud analysis methods for discontinuity identification, a rock mass discontinuity identification method based on an improved Density Peak Clustering (DPC) algorithm is proposed. Firstly, neighboring points are searched to estimate point cloud curvature and normal vectors, followed by the removal of high-curvature edge points. Secondly, a dual strategy of data sampling and data space gridding is employed to reduce the computational complexity of the DPC algorithm, where Gaussian kernel density estimation and density-distance calculation are performed on grid points. Thirdly, the optimal Gaussian kernel bandwidth <italic>h</italic>_opt and the optimal number of clusters <italic>M</italic> are determined through cross-validation and exhaustive search, and clustering labels are assigned to all points based on the squared sine distance. Finally, the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, in conjunction with Principal Component Analysis (PCA), is employed to adaptively segment discontinuities and to determine their orientations. The reliability of the proposed method was validated using regular-shaped point cloud and rock slope point cloud from the publicly available Rockbench repository, and the method was further successfully applied to a hazardous rock cut slope in China. The results demonstrate that, compared with other discontinuity identification methods, the proposed approach offers significant advantages in terms of accuracy, adaptability, and computational efficiency, making it highly suitable for processing large-scale rock mass point cloud data. The research outcomes provide an objective and efficient intelligent measurement tool for large-scale discontinuity surveys in engineering practice.