Long-term displacement measurement of full-scale bridges using camera ego-motion compensation

Civil infrastructures experience long-term deflection as a result of persistent and transitory loadings, including self-weight, pre-stress, traffic, temperature variation, and stress redistributions due to damage. Thus, long-term displacement is an important safety indicator that can be widely employed in structural health monitoring. However, its measurement is challenging because of the errors induced by the ego-motion of sensors. Researchers have attempted to compensate for the motion-induced error in computer vision-based displacement measurement methods for full-scale civil structures by using fixed objects in the background. Because remotely located objects cannot fully provide six degrees-of-freedom (6-DOF) camera motions, further developments are necessary for complete error compensation. This paper proposes a long-term displacement measurement strategy that uses computer vision-based ego-motion compensation. The proposed system consists of main and sub-cameras attached to each other. While the main camera employs the conventional computer vision-based method for displacement measurement, the sub-camera measures the ego-motions of the dual-camera system from which the motion-induced errors are estimated and compensated for. The proposed long-term displacement measurement algorithm was numerically validated, and the sub-camera was found to provide a noticeable error compensation. A laboratory-scale test showed that the motion-induced error was reduced from 44.1 mm to 1.1 mm. A field application conducted upon a newly constructed railway bridge provided continuous long-term displacement measurement data, which were consistent with LiDAR-based displacement measurements and numerical predictions.