Abstract: Dual-arm cooperative robot system has advantages of high efficiency, large load and strong collaborative ability. However, the performance and quality of dual-arm operation are affected not only by the positioning accuracy of a single arm but also by the collaborative positioning accuracy of dual arms, so a kinematic calibration method based on the combination of parametric and non-parametric models is proposed. Firstly, the kinematic model and parameter error model of the robot are established based on the MDH (modified Denavit-Hartenberg) method, the coupling parameters in the error model are removed, and the geometric parameter error is identified based on the iterative least square method. Secondly, since the traditional non-geometric error compensation method can only establish the mapping relationship between the joint position and the end-effector position error in the calibration coordinate system, an improved non-geometric error compensation method is proposed to compensate the non-geometric error of the robot body. Thirdly, the parameters of the transformation matrix of the dual-arm base coordinate system are identified based on the distance error, and the geometric and non-geometric errors of the dual arms are compensated. Finally, the correctness and effectiveness of the method are verified by experiments. The results show that the average positioning errors of UR10 and UR5 robots are reduced to 0.170 9 mm and 0.050 9 mm by the proposed method. The average cooperative positioning error of the dual arms is reduced to 0.167 6 mm, and the accuracy of the proposed method is improved by 27.7% compared with the method based on the parametric model, which verifies the superiority of the proposed method.