High-fidelity bridge damage assessment via enhanced 3D Gaussian splatting

Liu, Jiucai ORCID: https://orcid.org/0009-0001-7056-8983, Wei, Jiaqi, Xiong, Guanyu, Li, Haijiang ORCID: https://orcid.org/0000-0001-6326-8133, Pan, Yue and Wang, Dalei 2026. High-fidelity bridge damage assessment via enhanced 3D Gaussian splatting. Necati Catbas, F., Frangopol, D.M. and Yun, H-B., eds. Moving Toward Smart, Resilient and Sustainable Bridges, London: CRC Press, pp. 1822-1829. (10.1201/9781003778677-220)

Full text not available from this repository.

Abstract

High-fidelity bridge damage assessment requires accurate spatial localization, detailed surface texture representation, and effective organization of inspection information. Existing image- and point-cloud-based approaches often address these requirements only partially, limiting their applicability to large-scale bridges. Recent advances in 3D Gaussian Splatting (3DGS) provide an efficient neural rendering paradigm for photorealistic reconstruction, yet standard pipelines are not optimized for damage-oriented tasks. This paper presents an enhanced 3DGS approach for high-fidelity bridge damage assessment. To preserve damage-related surface details at scale, a block-wise, texture-oriented densification strategy selectively enhances Gaussian representation in texture-rich regions. Component- and defect-level segmentation results are embedded into a graph-based semantic representation, enabling structured organization, semantic querying, and assessment-oriented visualization. Experiments on a real-world bridge demonstrate improved defect visibility, consistent reconstruction, and actionable semantic-guided damage rendering, facilitating practical damage assessment and maintenance planning.

Item Type:	Book Section
Date Type:	Publication
Status:	Published
Schools:	Schools > Engineering
Publisher:	CRC Press
ISBN:	9781003778677
Last Modified:	02 Sep 2026 11:30
URI:	https://orca.cardiff.ac.uk/id/eprint/189350

Actions (repository staff only)

Edit Item