Abstract Summary
The increasing slenderness and lightness of modern footbridges make the vibration serviceability assessment a key aspect for their design and maintenance. Vibration assessment of footbridges relies on an in-depth knowledge of the footbridge dynamic behavior as well as on reliable models of human-induced loads. A promising approach to the characterization of footbridge dynamic behavior is represented by computer vision-based techniques. Over the past few years, computer vision-based monitoring has acquired increasing importance thanks to the availability of high resolution and frame rate consumer-grade cameras. In contrast to traditional monitoring systems relied on dense sensor networks, computer vision-based monitoring requires the installation of one or more cameras from different measuring points together with, if necessary, some targets on the monitored structure. The high performance of consumer-grade cameras and the contactless and non-destructive nature of the computer vision approach can make this technology particularly suited to the dynamic monitoring of footbridges. Nevertheless, the accuracy that can be reached in terms of displacement measurements and identified modal properties needs to be investigated. In this context, the paper presents the short-term dynamic monitoring of a steel footbridge based on computer vision techniques. An action camera was adopted to acquire videos with 4k resolution and 60 fps (frame per second). The structural deflection caused by both a jumping pedestrian and the wind were recorded from different measurement positions. The post-processing of the video frames (image processing, noise filtering, …) is presented and discussed in the paper. Special attention is paid to the use of circular targets placed on the footbridge, which allowed for the identification of vertical deflections smaller than one pixel. A traditional accelerometer-based monitoring system is also installed on the footbridge for validation purposes. Displacements evaluated through a double integration of the measured accelerations are compared to those obtained from the image processing. Results demonstrate the high potential of computer vision-based systems for the monitoring of structures and infrastructures.
