Algorithm for UAV flight controlling along a railway using technical vision

The paper proposes an algorithm for controlling the flight of an autonomous unmanned aerial vehicle (UAV) along the railway without operator participation and the use of satellite navigation systems such as GPS and GLONASS. With no operator and satellite navigation systems being involved, it becomes possible to inspect tracks over long distances, as the communication range with the UAV and external electromagnetic interference are of no concern. The algorithm is based on the use of computer vision: determining the location of rails in the video image and generating control signals to control the pitch, yaw and roll of the UAV in such a way as to keep the image of the rails in the middle of the video frame. Experimental UAV flights showed that the algorithm reliably determines the rails position and keeps the UAV flying along the railway.