Real-Time Vision-Based Dynamic Monitoring of Vertical and Horizontal Transverse Displacements in Bridge Decks

The objective of this study is to present a vision-based system that can extract in real-time displacements at multiple cross sections of bridge decks, which is accurate, easily replicable, and cost-effective. The key feature to accomplish such a goal is the reduction of the computational time required for displacement extraction, achieved through the use of multiple image subdomains whose position is continuously updated in the acquired video footage, following the movements of the monitored targets. This optimization at the software level is paired with choices in the hardware configuration that provide major benefits, namely: (1) camera below the bridge deck with its optical axis parallel to the bridge axis to obtain vertical and horizontal transverse displacements of any cross section, even with a single video camera; and (2) use of coded targets for automated camera calibration, automated definition of the initial positions of the subdomains, and automated compensation of the vibrations of the video camera employing a reference point in the abutment or intermediate pier that naturally falls within the camera field of view. An application to a post-tensioned concrete bridge during normal traffic conditions is used to analyze the quality of the obtained results through comparisons with contact sensors (displacement transducer and accelerometer). The results show that the proposed computer vision system meets the declared research objectives, given its precision, even in the case of displacements of a few millimeters, as in the considered case study, broad applicability for the most common bridge decks, and the optimized use of affordable hardware.