A conditional diffusion model for bridge point cloud repair

Tools

Fang, Yunping and Ninic, Jelena; Moreno-Rangel, Alejandro and Kumar, Bimal, eds. (2025) A conditional diffusion model for bridge point cloud repair. In: EG-ICE 2025. University of Strathclyde Publishing, GBR, pp. 359-368. ISBN 9781914241826 (https://doi.org/10.17868/strath.00093264)

[thumbnail of Fang-Ninic-EG-ICE-2025-A-conditional-diffusion-model-for-bridge-point]
Preview
 Text. Filename: Fang-Ninic-EG-ICE-2025-A-conditional-diffusion-model-for-bridge-point.pdf
Final Published Version
License: Creative Commons Attribution 4.0 logo
Download (1MB)| Preview

Abstract

Accurate and complete point cloud data are essential for the digital reconstruction of large-scale infrastructure. However, point clouds acquired in real-world scenarios are often incomplete due to occlusions, sensor limitations, and environmental constraints. This study extends a conditional denoising diffusion probabilistic model to address bridge point cloud completion. A point cloud encoder is designed to extract latent shape representations from incomplete inputs, which are then used to guide the generative process of the diffusion model. The method enables the production of high-fidelity, structurally consistent, and resolution-flexible point clouds. In addition, a dataset generation strategy is introduced to simulate typical point cloud defects encountered during scanning. Experiments on real-world bridge data validates the effectiveness of the proposed approach in addressing complex and large-scale point cloud completion tasks. Compared to the baseline, the modified encoder achieves consistent improvements across multiple structural components, with Chamfer Distance (CD) reduced by up to 23.2% and Earth Mover’s Distance (EMD) by up to 54.0%, indicating enhanced geometric accuracy and structural integrity.

  • Item type:Book Section
    ID code:93264
    Dates:
    Date
    Event
    1 July 2025
    Published
    9 June 2025
    Accepted
    Subjects:Fine Arts > Architecture
    Department:Faculty of Engineering > Architecture
    Depositing user: Pure Administrator
    Date deposited:27 Jun 2025 10:53
    Last modified:30 May 2026 09:44
    URI:https://strathprints.strath.ac.uk/id/eprint/93264
CORE (COnnecting REpositories)