Collision avoidance algorithms for unmanned aerial vehicles using computer vision

Abstract

We present a computer-vision-based approach to enable unmanned aerial vehicles (UAVs) to avoid collisions. To detect a moving obstacle, a machine learning algorithm, cascade classification, is used. Next, to track the obstacle, the camshift algorithm is implemented. These algorithms determine the coordinates of center of an object moving towards the UAV. Finally, to determine the distance of the object to the UAV, a stereo camera is used. Once an object is successfully tracked, the UAV will execute avoidance maneuvers if the path of the moving object conflicts with the UAV's path.