The mathematical model of a stereo vision sensor is analyzed. Aiming at the application of vision measurement in large Field of View (FOV), a novel calibration method for stereo vision sensor with large FOV is proposed. A baseline rulur is randomly placed in FOV of the vision sensor for several times and the images of the feature points on the baseline rulur are acquired by the cameras. Then, based on the foundational matrix and the constraint of the known distance between two feature points on the target, the intrinsic parameters of cameras and the structure parameters of stereo vision sensor are simultaneously estimated by the method combining linear solution with nonlinear optimization. The experimental results show that the proposed calibration method is efficient and easy to operate . All parameters of stereo vision sensor can be determined without initial value. By the measurement precision of 0.06 mm in the measurement range of 6 000 mm×4 500 mm,it is proved that the proposed method is suitable for calibrating stereo vision sensor with large FOV on the spot.