Algorithms of Runway Detection in Video Images in Neural Network Based UAV Landing

Algorithms for detecting the runway contours in video images based on YOLOv8 neural network are presented, differing by the types of problems it is trained to (detection, segmentation, pose estimation). The accuracy and speed of these algorithms run on NVIDIA Jetson NANO computer module are analyzed. Using the analysis results, the best detection algorithm is selected based on certain parameters (speed, accuracy, range). The results confirm that the algorithm can be applied in onboard software of unmanned aerial vehicles (UAV).