This paper presents and comparison two stereo vision sensor calibration methods: Volume template based calibration method and the calibration based on virtual stereo pattern. First method use a premade volume template. Second method use a group of random spatial points that are formed by an infrared LED moved with the probe of CMM. Two methods adopt an ideal pin-hole model that ignores the nonlinear distortion of the cameras. Through compare two calibratioin methods, we can draw conclusion that the first method can be applide to stereo sensor calibration in measurement field and the second method could gain higher calibrating precision.