Obstacle Detection for Unmanned Ground Vehicle in Cross-Country Environment

Reliable detection of non-traversable hazards is a key requirement and significant challenge to unmanned ground vehicle (UGV) autonomous cross-country navigation. DLUT is developing a daytime obstacle detection capability using UGV mounted sensors and has performed three obstacle detection algorithms. In this paper, we currently focus on the cues for obstacle detection color camera sensor provides. Color or monochrome images were collected from a general dynamics autonomous system and processed off-line with run-time passive perception software that includes: a water obstacle detector, a non-traversable tree trunk detector, a bush or shrub detector. A detailed description of each obstacle detection algorithm and their performance is presented in this paper.