Stereo vision based traversable region detection for mobile robots using u-v-disparity

Abstract

Abstract: Accurate and real-time traversable region detection is of great significance to the navigation of mobile robots. In Comparison with lasers and radars, stereo vision can provide richer information and thus has become increasingly popular in this field. In this paper, an efficient traversable region detection algorithm is proposed based on stereo vision by using u-v-disparity. In the algorithm, obstacle disparity image and non-obstacle disparity image are introduced to enhance the performance. The thorough experiments are performed on our NuBot rescue robot in the indoor RoboCup rescue field and the outdoor campus environment and on KITTI Vision Benchmark dataset, and the results show that the traversable region can be detected efficiently with high successful rate by using the proposed algorithm.