A two-stage iterative compensation method is proposed to solve the problem of the conversion error between the measurement coordinate system and the tool coordinate system in the existing robot kinematics calibration methods. The first level of compensation uses the linear search criterion to improve the Gauss-Newton method, uses the improved Gauss-Newton method to get the rough solution, initially calibrates the measurement coordinate system and the tool coordinate system, and then uses the least square method to get the detailed solution of the initial calibration value and the robot body, and carries out the complete kinematics parameter calibration compensation. The secondary compensation compensates the residual angle errors of the robot by the inverse kinematic iteration method. KR210_2700 robot is compensated by laser tracker. The experimental results show that after two times of compensation, the average absolute positioning error of the robot is reduced from 2.862 to 0.571 mm, and the average positioning accuracy is improved by about 80.05%, which effectively improves the absolute positioning accuracy of the robot.