Using vision technique for the bridge deformation detection

A three-dimensional photogrammetric-measurement vision system has been developed which consists of TV cameras, a PC/AT computer, 3-D control points and software for both computer vision and photogrammetry. The bulk of the processing is the determination of the subpixel location of the control point in the stereo images and accurate camera calibration, using direct-linear-transformation (DLT) and error-approximation algorithms. Simulation experiments which apply this system to bridge deformation detection are carried out. The average 3-D measurement error (x, y, z):(dx, dy, dz)=(0.24, 0.17, 0.24) mm, or 1/2000 over the field of view and 1/3500 over the depth. The bridge deformation measuring accuracy is ±0.323 mm