An original approach for the patrol robot in the substation to detect the obstacle height is presented based on binocular stereovision. The road model hypothesis of the robot’s environment is proposed. The distance between the obstacle and the camera is computed on the basis of the algorithm of binocular distance measurement. With the selected sample in the disparity image

the road plane equation of the current frame is calculated by the improved random sample consensus（ RANSAC） algorithm and estimated and predicted by the Kalman filter. When the distance between the obstacle and the road is calculated on the basis of the formula from the point to the plane

the height image is obtained. The obstacle is detected according to the height information. Applications show that the algorithm is robust to the shadows and illumination change and more adaptable to the environment. The obstacle height is precise enough to meet the requirement.