Vision-Based Methodology for Characterizing the Flow of a High-Density Crowd on Footbridges: Strategy and Application

Abstract

Obtaining pedestrian trajectories by a vision-based methodology is receiving increasing attention in the literature over recent decades. Within the field of study of human-induced vibrations on footbridges, practical challenges arise when collecting the trajectories of high-density crowds during measurement campaigns. A cheap and robust methodology tackling these issues is presented and applied on a case study consisting of a real-life footbridge occupied with many pedestrians. A static camera setup consisting of low-cost action cameras with limited installation height is used. In addition, a drone camera was employed to collect a limited amount of footage. Pedestrians are equipped with colored hats and detected using a straightforward color-segmenting approach. The measurements are subjected to both systematic and random measurement errors. The influence of the former is theoretically investigated and is found to be of limited importance. The effect of the latter is minimized using a Kalman filter and smoother. A thorough assessment of the accuracy results reveals that the remaining uncertainty is in the order of magnitude of 2 to 3 cm, which is largely sufficient for the envisaged purpose. Although the methodology is applied on a specific case study in the present work, the conclusions regarding the obtained accuracy and employability are generic since the measurement setup can be extended to a footbridge with virtually any length. Moreover, the empirically obtained results of the presented case study should find use in the calibration of pedestrian dynamic models that describe the flow of high-density crowds on footbridges and the further development of load models describing crowd-induced loading.

Eeklo Footbridge: Benchmark Dataset on Pedestrian-Induced Vibrations

“…The exact location and orientation of each camera was determined through bundle adjustment using a set of two-dimensional (2D)-3D correspondences of a total of 331 marker points on the bridge deck and parapet. The world coordinates of these marker points were obtained using a total station (standard deviation measurement error, 1 mm) (Van Hauwermeiren et al 2020).…”

“…All 21 cameras were synchronized offline based on common events captured in adjacent cameras. The video footage was then used to obtain pedestrian trajectories through a sequence of pedestrian detection and tracking, as comprehensively described in Van Hauwermeiren et al (2020). First, the pedestrians were detected in the consecutive frames using a color-segmenting approach.…”

HIT HAR: Human Image Threshing Machine for Human Activity Recognition Using Deep Learning Models

“…The classification results from the HIT machine show that the proposed HAR approach is a valid method for healthcare applications, including abnormal activity detection, elderly care in homes, and disabled assistance. The extended versions of HIT machines are helpful in other applications, including intelligent environments, indoor navigation [ 99 ], security and surveillance, and people monitoring [ 100 ].…”

Dynamic running actions on footbridges: A pilot study on human-structure interaction

“…The footbridge considered in these measurements is located in Eeklo, Ghent and is frequently used by cy-clist, joggers and pedestrians to cross the N49 road (Van Nimmen et al 2014). The footbridge consists of one main (arc-shaped) and two side (straight) beams with a total span of 96 meters and a total width of 3 meter (Van Hauwermeiren et al 2020). A more comprehensive and detailed description of the footbridge can be found in (Van Nimmen et al 2014).…”