Vision-based UAV Positioning Method Assisted by Relative Attitude Classification

Abstract

When the Unmanned Aerial Vehicle(UAV) is flying in formation, the common communication method is radio frequency(RF) communication. However, in practical applications, the way of RF communication is susceptible to interference from other factors such as electromagnetism. Therefore, in order to improve the anti-interference of the UAV cluster flight, it's necessary to use a positioning method which is based on visual information. Based on the above analysis, this paper proposes a vision-based UAV positioning method assisted by attitude classification. Firstly, the problem of solving the relative attitude of the UAV is transformed into a classification problem by the object recognition method, and a preliminary classification of the relative attitude of the friendly UAV is realized. Based on the principle of camera calibration, the pixel size and coordinates of the target UAV can be transform to the body coordinate system. Since the camera and the carrier UAV are fixedly connected, when the latitude and longitude coordinates of the carrier UAV are known, relative coordinate conversion can be performed to calculate the coordinates of the target UAV in the world coordinate system. Realize the positioning task of the target UAV. Simulation results are performed on the proposed method of the UAV relative attitude recognition accuracy exceeds 90%, and the average error in the distance simulation system of 2.56%. The final coordinate positioning accuracy exceeds 90% without losing the target.