Aerial target detection and avoidance for UAV based on stereo vision

Aerial target detection and avoidance is vital for the flight safety of unmanned aerial vehicles (UAVs). The limited amount of feature points for small and mobile obstacles make them very difficult to be detected. In this paper, the background segmentation and small target detection algorithm is applied in the obstacle detection, and the obstacle localization is achieved by the triangulation method, after which, the threat of collision is evaluated using the safety envelope method. Based on the evaluation results, a UAV can avoid the obstacles by the target point avoidance method. Experiments using a six rotor equipped with a stereo vision system have been implemented to testify the method effectiveness and the results show that the UAV can detect and avoid obstacles effectively.