Distance Measurement Algorithm Based on Binocular Stereo Vision

Abstract

Distance measurement technology of binocular stereo vision has the advantages of wide-range detection, simplicity and reliability. The method is widely applied to robot obstacle avoidance and path planning. Binocular stereo vision can only measure the distance of images feature points generally. However, more information about distance of non-feature points is also needed to acquire in practical applications. This paper proposes a stereo distance measurement method, which can measure distance of points whether it is a feature point or not based on a dense matching method. A dense parallax map is obtained by the graph-cuts algorithm. On the basis of the calibration parameters of binocular camera and the left and right image dense parallax map, the three-dimensional coordinates of the any points and their distance will be gotten. The true image experiment has proved the feasibility of this algorithm with high accuracy and maneuverability.