Publications / 2026 Proceedings of the 43rd ISARC, Singapore

Automated Scan-to-BIM Method for City-Scale Building Roof Reconstruction Using UAV Photogrammetry and Geospatial Building Footprints

Boan Tao, Frédéric Bosché, Jiajun Li
Pages 1602-1609 (2026 Proceedings of the 43rd ISARC, Singapore, ISBN 978-0-6458322-3-5, ISSN 2413-5844)
Abstract:

The digital transformation of existing building stock through Building Information Modeling (BIM) remains a significant goal and challenge in the Architecture, Engineering, and Construction (AEC) industry. This paper presents an automated Scan-to-BIM framework specifically designed for IFC-compliant roof model generation from Unmanned Aerial Vehicle (UAV) photogrammetry data. The proposed method leverages geospatial building footprint data as semantic priors to intelligently filter large-scale 3D meshes, enabling targeted roof extraction and IFC modelling from large-scale meshes. A robust segmentation pipeline combining Taubin smoothing, region-growing algorithms, and adaptive height filtering successfully extracts planar roof surfaces. The extracted geometries are reconstructed into Industry Foundation Classes (IFC) compliant BIM objects using alpha-shape boundary regularisation, which preserves non-convex features such as conical spires and L-shaped roofs. Experimental validation is first conducted on a large building with a complex roof structure composed of 431 roof panels for which a roof patch segmentation precision of 98.8\% is achieved. The framework is subsequently demonstrated at the scale of a university campus to assess scalability and robustness.

Keywords: Scan-to-BIM, Semantic Segmentation, Industry Foundation Classes (IFC), Digital Twin, Automated Roof Extraction, Geospatial Data Fusion.