<p>Rockfall hazards pose a significant threat to infrastructure and human safety in mountainous regions, particularly in rock masses with densely developed discontinuities. Accurate localization of potential rockfall sources is essential for effective hazard assessment. In this study, a semi-automatic method is proposed for localizing rockfall sources by extracting discontinuity-controlled rock blocks from high-resolution point clouds. The main steps are as follows: (1) A simplified Clustering by Fast Search and Find of Density Peaks (CFSFDP) algorithm is introduced to efficiently identify discontinuity sets, and individual discontinuities are then extracted using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. (2) A semi-automated method is then developed to identify discontinuity combinations of multi-face exposed blocks characterized by Exposed Vertices (EVs). (3) Based on the identified discontinuity combinations, blocks are generated through spatial intersection, and geometrical properties, including volume and areas, are calculated. (4) The failure mode and stability analysis of the block are conducted using the block theory. Compared with the existing methods, the proposed method introduces a simple yet effective geometric criterion framework for determining valid discontinuity combinations and imposes no limitation on the number of discontinuity sets. Two case studies—a synthetic box model with well-defined planar features and a natural rock slope in a mining area—are used to validate the method. The results show that the method accurately reconstructs block geometries in controlled settings and exhibits reasonable performance under natural conditions. By integrating geometric characterization of discontinuities with mechanical stability analysis, the proposed method provides a practical and effective approach for identifying potential rockfall source zones from point cloud.</p><p><b>Highlights</b><UnorderedList Mark="Bullet"> <ItemContent> <p>A modified CFSFDP clustering algorithm is proposed for discontinuity set identification.</p> </ItemContent> <ItemContent> <p>A semi-automated method to extract rock blocks directly from point clouds based on the spatial relationships between discontinuities.</p> </ItemContent> <ItemContent> <p>Multi-face-exposed blocks are reconstructed, and their geometric and mechanical properties are evaluated.</p> </ItemContent> <ItemContent> <p>Failure mode and stability of each block support spatially explicit rockfall source identification.</p> </ItemContent> </UnorderedList></p>