In camera parameter calculation, many distortion models ignore the effect of Center of Distortion (COD) and take the principal point or the center of the image as COD. In order to improve the precision of calibration result, the COD is introduced to the distortion model and added to the intrinsic parameters. To compensate for the problems of large computation burden and low efficiency due to the added parameter, Stochastic Parallel Gradient Descent (SPGD) is introduced to optimize the result. Results of digital simulation and experiments show that the SPGD based multi-parameter camera calibration method has high precision, fine robustness and can converge rapidly.