Artificial Intelligence and Data Science
Online ISSN : 2435-9262
Fundamental study on FE model construction of a steel I-girder using SfM/MVS point clouds
Nanaka SAEKI, Masayuki TAI, Ryo WATANABE
Author information
JOURNAL OPEN ACCESS

2026 Volume 7 Issue 3 Pages 701-710

Details
Abstract

In recent years, the deterioration of social infrastructure, including bridges, has become a significant issue, increasing the demand for advanced maintenance and management methods. Among these methods, structural performance evaluation using three-dimensional point cloud data generated by Structure from Motion (SfM) and Multi-View Stereo (MVS) has attracted considerable attention. However, point clouds generated by SfM/MVS often contain missing data and outliers, and consist only of surface information, making it difficult to directly construct hexahedral finite element (FE) models with through-thickness discretization. In this study, a method for generating marker point clouds by triangulation using markers detected from video images is proposed. The generated marker point clouds are utilized for member surface estimation and point cloud reconstruction, and a hexahedral FE model is subsequently generated. To verify the applicability of the proposed method, a steel I-girder specimen was investigated. Comparison with results from a static loading test demonstrated that the constructed FE model was able to reasonably reproduce the stress distribution within the elastic range. The results indicate that the proposed approach can effectively support the construction of FE models from SfM/MVS point clouds and has potential for structural assessment of bridge components.

Content from these authors
© 2026 Japan Society of Civil Engineers
Previous article Next article
feedback
Top