Study of Lateral Displacements and the Natural Frequency of a Pedestrian Bridge Using Low-Cost Cameras

Vision-based techniques are frequently used to compute the dynamic deflections of bridges but they are rather computationally complicated and require demanding instrumentation. In this article, we show that it is possible to reconstruct the 2-D k11inematics of flexible bridges using a simplified algorithm to analyze common video imagery. The only requirements are that the movement of the control points is clearly visible on the images and that next to each control point, there exist vertical and horizontal bridge elements defining the image scale. We applied this technique during controlled, forced excitations of a timber bridge that was stiff in the vertical but very flexible in the lateral axis because of cumulated damage. We used videos from low-cost cameras, in which the changes of the pixel coordinates of several control points during excitation events and their attenuation were clear. These videos were obtained during two annual structural health monitoring surveys using numerous sensors (Global Navigation Satellite Systems (GNSS), robotic total station (RTS), accelerometers), and hence the output of the video analysis was fully controlled. Because of various errors, the transformation of the video image coordinates into bridge coordinates yielded spurious deflections along the main axis of the bridge, which were used to control the uncertainty of our results. We found that the computed lateral deflections (i) were statistically significant, (ii) satisfied structural constraints, and (iii) were consistent with structural estimates derived from other sensors. Additionally, they provided accurate estimates of the natural frequency and the damping factor of the bridge. This approach can be applied in other cases of monitoring of flexible structures if the requirements for planar deformation, pixel resolution and scale definition are satisfied.