LiDAR assisted 3D Gaussian splatting for surface deterioration mapping in semi-subterranean built heritage
Subterranean and semi-subterranean built heritage is difficult to document because low light, irregular masonry surfaces, constrained geometry, and moisture exposure complicate visual inspection and digital survey. This study develops a LiDAR-assisted 2D-to-3D Gaussian Splatting workflow for mapping surface deterioration in two historic werfkelders in Utrecht. The workflow combines image-based reconstruction, LiDAR-supported registration, CVAT and SAM 2-assisted annotation, and multi-view projection of curated 2D masks into point-based and 3DGS representations. Moisture-related and stress-related deterioration were examined. Results show meaningful spatial correspondence between the two representations, with broader moisture-related patterns transferring more consistently than fragmented crack-like patterns. Because both outputs derive from the same curated image masks, the quantitative comparison measures cross-modal consistency rather than independent classification accuracy. The workflow transforms image-derived surface-condition observations into spatialized evidence for heritage documentation, communication, and preliminary interpretation in challenging interiors.
Authors
- E. Verbree (ORCID: https://orcid.org/0000-0002-6892-6702)
- Steffen Nijhuis (ORCID: https://orcid.org/0000-0001-6066-822X)
- Yuyang Peng (ORCID: https://orcid.org/0009-0001-7665-9456)
- Zhuoyue Wang
- Rosina Petra Loman
- Vidushan Rajkumar Kandasamy
- Joep Leonardus Bisschop
- Timo Petrus Franciscus Michel Visser
- Luuk Lukas Bertus Johannes Fransen
- Yingwen Yu
Institutions
- Utrecht University (NL)
- Delft University of Technology (NL)
Publication Details
- Journal
- npj Heritage Science
- Published
- 2026-11-18
- DOI
- https://doi.org/10.1038/s40494-026-03090-6
- Primary Topic
- 3D Surveying and Cultural Heritage
- Type
- article
- Field-Weighted Citation Impact
- 0.00