As the traditional calibration methods are difficult to ensure the calibration accuracy and system stability, a new optimization design of the secondary calibration based on Levenberg- Marquard (LM) algorithm is proposed. It is the optimization and correction of the first calibration results. The method introduces the optimization idea of the LM algorithm. The required objective function can be obtained by capturing the image of standard block after the first system calibration. The problem is solved satisfactorily with the optimization theory of LM algorithm. Experimental results show the optimized system accuracy is less than 0.01mm, RMS error is less than 0.02mm. The proposed algorithm has significant effect on accuracy improvement, and the system is more stable.