Error analysis and compensation of binocular-stereo-vision measurement system

Measurement errors in binocular stereo vision are analyzed. It is proved that multi-stage calibration can efficiently reduce systematic errors due to depth of field. Furthermore, for difficulty in carry-out of multi-stage calibration, the compensation methods of errors are presented in this paper. First, using standard plane template, system calibration is completed. Then, moving the cameras to different depths, multiple views are taken and 3d coordinates of special points on template are calculated. Finally, error compensation model in depth is established with least square fitting. Experiment based on CMM indicates the relative error of measurement is reduced by 5.1% with the proposed method in this paper. This is of practical value in expanding measurement range in depth and improving measurement accuracy.