The building facade structure is one of the essential features to describe the architectural model. To solve the problems of the existing algorithms such as the long time to extract the building facade structure, the spatial projection error, and the missing point cloud after conversion, this paper proposes a building facade structure extraction method based on point cloud segmentation. This method uses single clustering segmentation to extract windows, and dilutes the main structure through a point cloud Downsampling algorithm to reduce the density and complexity of point clouds. Introduce standard vector algorithm to remove high-density main point cloud to edge point cloud. Finally, the improved 3D line segment detection algorithm traverses the edge point cloud to achieve feature extraction of building facade structures. Comparative experiments using self-testing point clouds and Semantic 3D publicly available data demonstrate that the proposed method effectively extracts building facade structures, enhances efficiency, and improves the filtering of disordered line segments. This method offers a new approach for structural processing in point cloud facade processing and model construction.