For the hand-eye calibration of the vision service robot, the traditional hand-eye calibration technology can’t be realized which because the service robot is independently developed and there is no teaching device to feed back the pose in-formation of the service robot in real time. In this paper, a hand-eye calibration method based on ROS (Robot Operating System) is proposed. In this method, ROS system is used to accurately control the arm of the service robot to rotate in different positions for many times. Meanwhile, the head camera of the service robot takes images of a fixed point in the scene. Then, the nonlinear equations were established according to the homography matrix of the two images and the position and pose information of the ROS system, and the accurate hand-eye relationship was optimized by the least square method. Finally, an experimental platform is built and the proposed hand-eye calibration method is verified. The experiment results show that the method is easy to operate, simple algorithm and correct result, which verifies the ef-fectiveness of the algorithm and provides conditions for the realization of humanoid grasping of visual service robot.