Semantic Segmentation of building point clouds based on Point Transformerand IFC

In the process of building construction, the semantic understanding of building point clouds provides potential solutions for efficient building quality supervision, progress monitoring, and sub-system deviation analysis.However, the lack of suitablepubliclabeled datasets, the low degree of color discrimination and the particularity of multi-system coexistencein construction scenes have brought certain challenges to semantic segmentation of point clouds.Butrich semantic information in related Industry Foundation Classes(IFC)can be used to synthesize effective labeled data.Our main contributions are as follows:1) We propose asynthetic conversion method from BIM model to point cloud, andconstructa synthetic dataset for construction scenesbased on IFC. 2)We segment the colorless point cloud of the construction sceneintofivetypes (IfcSlab, IfcBeam, IfcWall, IfcColumn, IfcDistributionFlowElement) with Point Transformerand use focal loss to improve the segmentation accuracy of small-area componentswith synthetic data for data enhancement.