HFR-Video-Based 3-D Software Sensor for Bridge Displacement Monitoring

In this letter, we propose a novel vision-based 3-D software sensor using high-frame-rate (HFR) videos for bridge measurement that can realize noncontact full-view 3-D dynamic analysis, including displacement distribution and frequency response. Model-based 3-D to 2-D projection is applied for addressing weak-texture bridge images to effciently match corresponding points in stereo HFR 4 K videos. Then, relative 3-D displacement is precisely calculated based on a pair of 2-D displacement vectors, which are estimated on each camera by digital image correlation with subpixel accuracy. 3-D displacement signals are analyzed by short-time Fourier transform for vibration analysis to calculate frequency responses of all measured points, and 3-D mode shapes of peak frequencies that refer to intrinsic structure characteristics are estimated in frequency domain. To verify its effectiveness, a prestressed concrete girder bridge of 23 m length in Japan was measured, and the experimental results show that our software sensor can achieve multipoint 3-D displacement measurement with an accuracy of 0.01 mm and 3-D vibration analysis in the frequency range of 0–50 Hz.