Research on Obstacle Recognition Methods of Unmanned Vehicles in Off-road Environment

The off-road environment is complex, and the accurate obstacle detection of unmanned vehicles in the off-road environment is difficult. In order to identify the obstacles in the off-road environment with low cost, real-time image recognition and high efficiency, the Faster R-CNN optimized by 4 methods is designed in this paper. Comparing with the original model, SSD and YOLO, our optimized model shows obvious advantages in obstacle identification in the off-road environment.