This paper focuses on the problem of the self-calibration of a stereo rig and the visual control method for 3-DOF approach movements of a robot. A new measurement model for the stereo rig fixed beside the robot is developed based on the relative position of the end-effector and target. The stereo rig is linearly calibrated online with at least two steps of unspecific motions of the robot. A mixed visual control method is then proposed to guide the end-effector to approach the target, whose robustness and efficiency are improved by taking the advantage of both position-based and image-based visual control methods. Experimental results are provided to verify the effectiveness of the proposed methods.