The problem of inaccurate initial parameters perplexed the Tsai two-step calibration method,hence a three-step calibration method based on linear projection constraints is proposed for camera calibration in order to improve the size detection precision of mechanical parts.Firstly,the linear projection constraint conditions were used to solve the initial accurate value of the image center and the distortion parameters.Combined the radial alignment constraint conditions with overdetermined equations,all the external parameters as well as the rest of internal parameters are then solved.Finally,the nonlinear optimization is conducted to all the parameters by constrained conditions.The world coordinate system based on image template is established to verify the effect of the proposed method.Using the image pixel coordinates,the camera coordinates are calculated and then normalized.The world coordinates corresponding to image point are solved using the external parameter matrix.An experiment is carried out with single CCD,and the coordinate results are comparatively analyzed with Tsai method and Zhang method.The experiment results show that this proposed method can effectively improve the accuracy of camera calibration,with pixel error reaching 0.126 5 and distance error of arbitrary two points in monocular visual size detection being less than 0.41%.