| Unmanned aerial vehicles(UAV) play a more and more important role in the modern battlefield. In order to improve the operational effectiveness of UAV, autonomous aerial refueling technology has been extensively studied by many countries in the world. Compared with traditional navigation methods, vision-based navigation method has the advantages of high precision, fast frequency and immunity to electronic interferences. This thesis is accomplished based on the study of vision-based navigation for UAV probe and drogue autonomous aerial refueling.Firstly, based on the characteristics of UAV probe and drogue autonomous aerial refueling, a design for vision-based navigation system is proposed, inlcuding the usage scenarios, composing and workflow of the vision-based navigation system.Secondly, vision-based navigation algorithms for probe and drogue autonomous aerial refueling are proposed in this thesis. Based on characteristics of auxiliary identification marks, image pre-processing algorithm and feature point extraction algorithm are proposed. The randomized Hough transform algorithm is employed for detecting the drogue, and the Kalman filter is used for tracking drogue. To solve the pose estimation problem for point features, traditional orthogonal iterative algorithm is studied. Based on the problems of the traditional algorithm, an improved algorithm is studied and adopted for estimating the pose between UAV and drogue.Thirdly, digital simulation system of vision-based navigation for probe and drogue autonomous aerial refueling is designed. In the digital simulation system, aerial refueling scenarios is simulated with Vega, control system is simulated with Simulink, and MFC connects Vega and Simulink to form a close simulation loop. The experimental results of the digital system show the effectiveness of the proposed vision-based navigation algorithms for UAV probe and drogue autonomous aerial refueling.Finally, semi-physical simulation system of vision-based navigation for probe and drogue autonomous aerial refueling is designed based on the digital simulation system. In the semi-physical simulation system, a 3D scenery simulation computer simulates the navigaton image acquired by the camera of the vision-based navigation system and projects the navigation image onto a curtain. A visual servo computer uses a real camera to capture the image projected on the curtain, and based on the captured image as well as the equivalent projection principle the relative pose between UAV and drogue in the virtual environment is calculated, UAV is then controlled to approach the drogue and finally accomplish the docking with drogue, which concludes the semi-physical simulation. The experimental results of the semi-physical system further show the effectiveness of the proposed vision-based navigation algorithms for UAV probe and drogue autonomous aerial refueling. |
