Stereo vision sensor calibration based on random spatial points given by CMM

This paper presents a new, high-precision calibration method for stereo vision sensor based on virtual template. Given 3D spatial points coordinates by coordinates measurement machine (CMM) and corresponding image points coordinates, the technique is realized by taking the projection matrix elements as unknown and using singular value decomposition to solve the least-squares solutions. The method avoids solving intrinsic and extrinsic parameters of each camera and introducing the calibration error which caused by template machining and measurement error. We measure the 3D coordinates of a group of random spatial points by the calibrated stereo vision sensor then compare the results with CMM measurement values, the errors are less than 0.05 mm in the Z axis direction, and less than 0.01 mm in the X and Y axis direction, the experiment results show that the technique is feasible and effective.