Computer vision-based 3D vibration displacement measurement and structural condition monitoring
This paper proposes a novel vision-based method for 3D structural dynamic displacement measurement using a binocular system free of on-site extrinsic calibration, integrated with a hierarchical nonlinear optimisation framework. Starting from only calibrated camera intrinsic parameters and a single known distance for metric scale, the unknown relative pose of the camera pair is recovered from a few shared scene features through sequential refinement of the fundamental matrix, extrinsic parameters, and 3D triangulation, all unified by reprojection error minimisation. A detection-descriptor decoupled strategy integrating SIFT keypoint localisation with DeDoDe-B learnt descriptors is introduced for robust frame-wise tracking with similarity match quality quantification. Experimental validations on rigid-body and flexible beam tests demonstrate superior accuracy over traditional DLT and KLT baseline, and the operational mode shapes of the aluminium beam are successfully identified.With additional mass added on the long beam, the proposed approach demonstrates the potential to identify the existence and position of the mass for structural condition monitoring.
Authors
- Xueren Wang (ORCID: https://orcid.org/0000-0002-8421-7279)
- Jun Li (ORCID: https://orcid.org/0000-0002-0148-0419)
- Hong Hao
- Dong Tan
Publication Details
- Journal
- International Journal of Structural Stability and Dynamics
- Published
- 2026-11-19
- DOI
- https://doi.org/10.1142/s0219455427410136
- Primary Topic
- Structural Health Monitoring Techniques
- Type
- article
- Field-Weighted Citation Impact
- 0.00