Research on Vision-based Autonomous Landing of Unmanned Aerial Vehicle

Abstract

In order to ensure that the Unmanned Aerial Vehicle (UAV) achieves a safe and accurate landing under special conditions such as GPS signal interference, an autonomous landing method based on computer vision is proposed. Firstly, the projection model of the airborne camera is established, and the camera’s internal parameter matrix is obtained by camera calibration. Then, the “H” landing target pattern is designed to reduce the complexity of image processing, and the white backing disc is added to make the background easier to the segment. Then the pose estimation model is established and combined with the coordinate information of the feature points to complete the pose calculation. Finally, the actual flight data of the UAV is compared with the pose estimation data, and the maximum error of the position and attitude estimation values is 0. 52m and 6.22°, which can basically meet the landing requirements of the UAV, and the feasibility of the method is proved.