This study presents a comprehensive investigation of methods and technologies for developing a computer vision-based framework for Structural Health Monitoring (SHM) and Structural Identification (St-Id) for civil infrastructure systems, with particular emphasis on various types of bridges. SHM is implemented on various structures over the last two decades, yet, there are some issues such as considerable cost, field implementation time and excessive labor needs for the instrumentation of sensors, cable wiring work and possible interruptions during implementation. These issues make it only viable when major investments for SHM are warranted for decision making. For other cases, there needs to be a practical and effective solution, which computer-vision based framework can be a viable alternative. Computer vision based SHM has been explored over the last decade. Unlike most of the vision-based structural identification studies and practices, which focus either on structural input (vehicle location) estimation or on structural output (structural displacement and strain responses) estimation, the proposed framework combines the vision-based structural input and the structural output from non-contact sensors to overcome the limitations given above. First, this study develops a series of computer vision-based displacement measurement methods for structural response (structural output) monitoring which can be applied to different infrastructures such as grandstands, stadiums, towers, footbridges, small/medium span concrete bridges, railway bridges, and long span bridges, and under different loading cases such as human crowd, pedestrians, wind, vehicle, etc. Structural behavior, modal properties, load carrying capacities, structural serviceability and performance are investigated using vision-based methods and validated by comparing with conventional SHM approaches. In this study, some of the most famous landmark structures such as long span bridges are utilized as case studies. This study also investigated the serviceability status of structures by using computer vision-based methods. Subsequently, issues and considerations for computer vision-based measurement in field application are discussed and recommendations are provided for better results. This study also proposes a robust vision-based method for displacement measurement using spatio-temporal context learning and Taylor approximation to overcome the difficulties of vision-based monitoring under adverse environmental factors such as fog and illumination change. In addition, it is shown that the external load distribution on structures (structural input) can be estimated by using visual tracking, and afterward load rating of a bridge can be determined by using the load distribution factors extracted from computer vision-based methods. By combining the structural input and output results, the unit influence line (UIL) of structures are extracted during daily traffic just using cameras from which the external loads can be estimated by using just cameras and extracted UIL. Finally, the condition assessment at global structural level can be achieved using the structural input and output, both obtained from computer vision approaches, would give a normalized response irrespective of the type and/or load configurations of the vehicles or human loads.
If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu.
Doctor of Philosophy (Ph.D.)
College of Engineering and Computer Science
Civil, Environmental, and Construction Engineering
Length of Campus-only Access
Doctoral Dissertation (Open Access)
Dong, Chuanzhi, "Investigation of Computer Vision Concepts and Methods for Structural Health Monitoring and Identification Applications" (2019). Electronic Theses and Dissertations, 2004-2019. 6867.