Liu, Jiucai  ORCID: https://orcid.org/0009-0001-7056-8983, Xiong, Guanyu, Wang, Dalei, Chai, Chengzhang ORCID: https://orcid.org/0000-0001-6911-8048, Pan, Yue and Li, Haijiang ORCID: https://orcid.org/0000-0001-6326-8133
2026.
Large-scale Gaussian splatting with semantic graph modeling for fine-grained bridge surface damage management.
Automation in Construction
191
, 107128.
10.1016/j.autcon.2026.107128
|
Preview |
PDF
- Accepted Post-Print Version
Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (20MB) | Preview |
Abstract
High-fidelity 3D reconstruction with precise defect representation is crucial for bridge inspection but remains challenging. Image-based methods often lack spatial awareness, while point clouds fail to capture fine surface textures. Standard Gaussian Splatting also struggles with large-scale reconstruction and detailed defect modeling due to global optimization and densification limitations. This paper presents a Gaussian Splatting framework for photorealistic and defect-sensitive reconstruction of large bridge structures with semantic embedding and hierarchical modeling. Panoramic imaging is adopted for initialization to ensure comprehensive coverage. A block-wise defect-aware densification strategy is introduced to selectively enhance Gaussian density in texture-rich regions while maintaining computational efficiency at scale. In addition, component and corrosion segmentation are incorporated into the Gaussian representation and organized through a graph-based structural model, enabling semantic querying and structured analysis of defect instances. Experiments on real-world bridge datasets demonstrate improved texture fidelity, more accurate defect representation, and enhanced semantic accessibility for inspection and maintenance tasks.
| Item Type: | Article |
|---|---|
| Date Type: | Publication |
| Status: | Published |
| Schools: | Schools > Engineering |
| Additional Information: | Rights Retention Policy applied |
| Publisher: | Elsevier |
| ISSN: | 0926-5805 |
| Date of First Compliant Deposit: | 10 September 2026 |
| Date of Acceptance: | 4 July 2026 |
| Last Modified: | 10 Sep 2026 10:30 |
| URI: | https://orca.cardiff.ac.uk/id/eprint/188812 |
Actions (repository staff only)
 |
Edit Item |
Dimensions
Dimensions