The vibration comfort evaluation is a control standard other than strength and deflection, but the general comfort evaluation method only considers the response of the mid-span position and does not consider the difference in the vibration response of different positions at the same time. It is crucial to study how pedestrians actually feel when they walk on footbridges. The computer vision-based vibration comfort evaluation method is a novel method with advantages, such as noncontact and long-distance. In this study, a computer vision-based method was used to evaluate the global vibration comfort of footbridges under human-induced excitation. The improved Lucas-Kanade optical flow method is used for multitarget displacement identification of footbridges. Additionally, the YOLOv5 algorithm for pedestrian detection is used to obtain the position information of pedestrians on the footbridges. Then, according to the pedestrian position information, the structural responses of different pedestrian positions corresponding to time periods are extracted from the displacement responses of each point, and they are combined to obtain the structural global displacement. The global acceleration can be obtained by calculating the global displacement. The rms value can be calculated based on the global acceleration and compared with the standard for comfort evaluation. The global comfort evaluation method is validated by pedestrian walking experiments with different frequencies on a laboratory footbridge. The experimental results show that the computer vision-based global comfort evaluation method for footbridges is feasible and is a more specific and real-time comfort evaluation method.
Keywords: computer vision; footbridge; pedestrian detection; pedestrian load; vibration comfort.