Camera calibration method based on 3D board

Jie Zhang,Xinde Li,Xianzhong Dai

DOI: https://doi.org/10.3969/j.issn.1001-0505.2011.03.022

2011-01-01

Abstract:A method of camera calibration based on 3D calibration board, which is used for vision measuring system, is proposed in this paper. Transform matrix is worked out with least square method on an ideal pin-hole camera model and then optimized by Levenberg-Marquardt(LM) algorithm. According to several transform matrices, initial camera intrinsic parameter matrix and corresponding extrinsic parameter matrices are solved. Then, the distortion coefficient is estimated according to the relationship between the real image and the detected image coordinate with initial camera parameter fixed. At last, LM algorithm is applied to the global optimization and the more accurate intrinsic and extrinsic parameters are obtained. Experimental results indicate that error of equivalent focal length is less than 0.1% and error of principal point is less than 0.5 pixel when Gaussian noise with less than 0.5 pixel standard deviation is added to the simulation image data. Under the same Gaussian noise level the more the pictures used, the smaller the standard deviation of camera parameters. Compared with other calibration methods 3D calibration board proposed in this paper is more accurate. More stable camera parameters can be obtained by using more pictures of the calibration board.