Stereo Vision Based Obstacle Avoidance Path-Planning for Cross-Country Intelligent Vehicle

Intelligent vehicle navigation in complex cross-country terrain can benefit from accurate obstacle avoidance path-planning. The paper presents an obstacle avoidance path-planning algorithm based on stereo vision system. Firstly, Multi-feature extraction based stereo matching techniques is proposed on a corner and edge extracted image pair to accomplish environmental perception, which can provide robust detection results under complex conditions. Secondly, in order to plan a path for a vehicle, a method of analyzing three-dimensional data produced by stereo vision is described. It computes estimates of interpolated height, slope, and roughness at equally spaced grids, as well as accuracy estimates of each grid. Then fuzzy rule is designed to integrate slope, roughness and certain value to analysis the traversability. To realize obstacle avoidance path-planning for cross-country intelligent vehicle, traversability is considered in Vector Field Histogram (VFH) algorithm and obstacle avoidance behaviors are designed. Experiments confirm the effectiveness of the obstacle avoidance path-planning method.