Abstract: The traditional two-stage calibration technique is improved based on the analysis of the principal calibration methods of robot vision. The technique firstly makes use of two-stage method to solve the initial value of the transition matrix,then establishes the optimization mathematical model,and Gauss-Newton method is used in the optimization computation. The proposed technique not only possesses as many virtues as the traditional two-stage method,for example,automationc,high speed,but also completely eliminates the accumulative error existing in the second stage of the traditional method and improves the calibration accuracy greatly. Many experiments indicate that the proposed technique excels the traditional two-stage method distinctly.