Stereo Vision-Based Obstacle Detection Using Fusion Method of Road Scenes

Free space and on-road obstacle detection is one of the key functions for the implementation of the vision-based intelligent vehicle and robot navigation system. Stereo vision-based algorithm for this task is more realistic and precious compared with radar or lidar-based algorithms. In addition, accurate estimation results can indicate the current and approaching conditions in the complex traffic scenes. A fusion method which combines polar occupancy grid and probability model for free space detection is proposed in this paper. The spatial and temporal filter is used to get more accurate results. After that, an adaptive membership cost function is applied for obstacle estimation on the road. Experiment results show that our method achieves high stability in a variety of traffic environments.