Aligning as-built and as-designed: Local point cloud to BIM registration via hybrid visibility map encoding for construction digital twins

Aligning local as-built data including point clouds and images with as-designed BIMs is critical for enabling construction digital twins. However, automated registration of local scan data to BIMs remains challenging due to modality gaps, large search spaces, structural self-similarity, and geometric inconsistencies. This paper leverages visibility maps for their inherent robustness to modality differences and strong discriminative capability in self-similar environments, proposing a hybrid visibility map encoding approach that integrates traditional geometric descriptors with deep features extracted from vision foundation models for local point cloud to BIM registration. Experiments on real construction sites showed over 92 % registration success, outperforming traditional local and global feature-based methods. The results enable more effective quality control, progress tracking, and model updating for stakeholders concerned with construction projects. This work paves the way for future research on applying foundation models to cross-domain data alignment and broader applications in the built environment sector.