Abstract: When there are errors in camera external parameters calibration and robot forward kinematics, the calibration results of robot-world coordinate system and hand-eye relationship can't converge to the global optimum. To solve this problem, a dual quaternion theory based method for simultaneous calibration of the robot-world and hand-eye relation is proposed. Firstly, the coordinate rigid transformations in calibration equation are parameterized by the screw axis, rotation angle and translation. Secondly, the translation is optimized based on global optimization algorithm. A numerical simulation system of PUMA560 robot and a real experiment platform of industrial robot are developed, on which the proposed method are analyzed and compared with the classical quaternion and dual quaternion calibration methods. Both the simulation and experimental results show that the proposed method can get global optimal solution even without initial estimation or data selection before calibration, when there are errors in camera external parameters calibration and robot forward kinematics.