Drogue Detection and Location for UAV Autonomous Aerial Refueling Based on Deep Learning and Vision

Abstract

The accurate detection and location of the drogue under complex environment is an important issue in UAV (Unmanned Aerial Vehicle) autonomous aerial refueling. In this paper, a new drogue detection and location method based on deep learning and vision is proposed for this intractable problem. The method consists of two parts: drogue detection and drogue location. The well-trained Yolo (You only look once) model is established to detect the drogue in the image to obtain the parameters of the predicted bounding box. A small part of the entire image is selected for processing based on these parameters, then the position of the eight beacons on the drogue ring in the image can be obtained. Least-squares ellipse fitting is performed on these eight points in the image coordinate system to obtain the long semi-axis of the ellipse. Finally, monocular vision is used to measure the position of the drogue in camera coordinate system. The simulation results show that this method can not only correctly identify the drogue but also accurately locate it with a distance of 2.5m to 45m under complex environment.