Binocular Camera Parallel Calibration Method Based on Sub-pixel
Juan Du, Yongchao Tang · 2020
During the operation of binocular vision equipment, the accuracy of initial calibration will gradually decrease. The aging of hardware platform and the change of temperature and humidity will lead to the decline of camera calibration accuracy. Therefore, the binocular camera needs to be quickly calibrated several times. To solve this problem, a parallel calibration method of binocular camera based on sub-pixel corner detection is proposed. The precision of calibration is improved by selecting the most appropriate number of calibration images. The internal parameter matrix and distortion coefficient matrix are obtained by using the parallel calibration method of monocular camera. Furthermore, the rotation matrix, translation matrix and re-projection errors of binocular camera are obtained. In order to improve the calibration accuracy of left and right cameras, a sub-pixel corner detection method is proposed to extract the corners of chessboard. The experiment results show that the parallel thread calibration algorithm is 23.5% faster than the traditional method. This algorithm can effectively improve the speed of binocular camera calibration while ensuring the camera calibration accuracy.