Monitoring Bridge Vibrations Using Non-Contact Sensors: A Deep Learning Approach
Students should possess strong programming skills in Python and camera calibration and synchronization, and have taken a deep learning course.
Monitoring the vibrations of infrastructure, such as bridges, is a critical aspect of Structural Health Monitoring (SHM) [1] to ensure safety. Traditional methods rely on contact-based sensors, which, while effective, are costly and challenging to deploy and maintain, particularly for large structures. Moreover, most existing research focuses on planar bridge vibrations, often overlooking vertical movements caused by factors such as strong crosswinds.
With advancements in Computer Vision and Deep Learning, non-contact sensors like cameras and drones have emerged as promising alternatives, offering cost efficiency and ease of deployment. However, accurately measuring 3D bridge vibrations [2] using these sensors presents two key challenges: camera pose estimation [3] and object tracking [4] in 3D. To address these, a specialized dataset for camera pose estimation and 3D bridge vibrations is required. This dataset would serve as a valuable benchmark, filling a critical gap in the research community.
This MSc research project aims to develop a simulated dataset for camera pose estimation and bridge vibrations in a controlled laboratory environment. Data acquisition will involve tripod-mounted cameras, drones, and contact-based sensors on a toy bridge. During this phase, you will collaborate with a PhD student for camera calibration, synchronization, and data processing.
You will also review literature on deep learning models for camera pose estimation, object detection, and tracking. The collected data will be used to train and evaluate a deep learning model for vibration estimation using drones, comparing its performance against “ground truth” data from contact sensors and the vibration estimation using the fixed tripod cameras.
[1] Dong C-Z, Catbas FN. A review of computer vision–based structural health monitoring at local and global levels. Structural Health Monitoring. 2021;20(2):692-743. doi:10.1177/1475921720935585.
[2] Shao Y, Li L, Li J, et al. Computer vision based target-free 3D vibration displacement measurement of structures[J]. Engineering Structures, 2021, 246: 113040.
[3] Wang J, Rupprecht C, Novotny D. Posediffusion: Solving pose estimation via diffusion-aided bundle adjustment[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2023: 9773-9783.
[4] Ren H, Han S, Ding H, et al. Focus on details: Online multi-object tracking with diverse fine-grained representation[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2023: 11289-11298.