Indoor 3D point cloud reconstruction for scan-to-BIM automation

In today's urbanizing world, the demand for intelligent digital representations of indoor spaces is increasing, driving advancements in automated 3D modeling and data integration. Building Information Modeling (BIM) transforms spatial digitization by converting raw point clouds into actionable models, supporting applications such as building management and smart city planning. Existing reviews on scan-to-BIM often focus on specific stages or technologies, overlooking a comprehensive workflow perspective. This paper employs an integrated review approach, combining scientometric analysis with a systematic qualitative review of the entire scan-to-BIM framework. It examines methods for reconstructing and modeling both structured and unstructured indoor elements using traditional and recent deep learning approaches. The review identifies key methodologies, limitations, and automation challenges, data quality, and modeling complexity, and outlines future directions to enhance scan-to-BIM workflows. By providing a comprehensive overview, this study aims to advance the understanding of scan-to-BIM automation and its contribution to indoor 3D BIM modeling. • Comprehensive analysis of the scan-to-BIM workflow from raw point clouds to BIM modeling. • Integrated scientometric and systematic reviews highlighting key advances and challenges. • Critical evaluation of methods for reconstructing structured and unstructured indoor elements. • Comparative discussion of traditional and deep learning approaches across processing stages. • Identification of limitations and future research directions for scan-to-BIM automation.