Stereo vision and hand-eye calibration is one of the most important and fundamental tasks in visually guided robotic control tasks. The work presented in this paper aims to develop an accurate camera based stereo vision and hand-eye calibration algorithm that is robust enough to be used in a robotic arc welding system. The proposed calibration method allows the object reference frame to be directly related to the robots base co-ordinate frame without the use of expensive coordinate measuring devices. Experimental results have shown that the proposed method can achieve a positional error of less than 1 mm in the Cartesian space, which is acceptable in most arc welding applications.