Addressing the challenges in facade door and window detection, a point cloud-based method utilizing void detection and rule constraints has been proposed. This method involves cloth simulation filtering followed by plane segmentation using the Random Sample Consensus (RANSAC) algorithm to extract facade point clouds and perform coordinate transformations. A virtual point cloud containing door and window information is then generated through point cloud inversion, and door and window clusters are identified using morphological methods based on point clouds. Finally, rules are applied to restore doors and windows missing due to occlusions, yielding the final detection results. Experiments on multiple building point clouds demonstrate that the proposed method effectively detects doors and windows on building facades, overcoming the impacts of occlusions. The detection results achieved precision, recall, and F1 scores of over 93%, 97%, and 95% respectively, indicating the excellent performance of the proposed algorithm.