Visual Detection and Identification of UAV s by YOLO Networks

The ever-increasing number of drones, especially of smaller sizes, defines new demands on their detection and identification, as conventional means are often inadequate. One possible solution is visual detection, which can be performed using inexpensive cameras, but it is a computationally intensive task. This paper deals with the use of a special family of convo-lutional neural networks, YOLO, which have been successfully used in e.g. computer vision. The paper describes the utilization of this resource in the task of detection and identification in specific environments with drones and presents solutions in the form of the use of evolutionary computation for the needs of adjusting the network parameters. The design was experimentally verified with a real drone and subsequently, the results were evaluated.