无人机位姿测量的点特征视觉方法

针对无人机在动态环境下快速高精度定位的问题，提出了用单目相机对无人机上的人工特征点进行位姿解算的方法。在无人机上放置一定数量的小型LED灯，并将其作为视觉测量的特征点，并以其中一个点作为原点建立无人机机体坐标系。通过多场景测量确定特征点在机体坐标系下的三维位置，再将三维位置与特征点在图像中的成像位置相匹配，最后使用EpnP算法求解出无人机的位置和姿态。在实验部分，利用三轴移动平台和三维转台，分别对位置解算结果和姿态解算结果进行误差测量。试验结果表明，位置解算误差在2%以下，姿态误差在8%左右。同时，该算法的处理时间在2 ms左右，该算法可以满足无人机对定位的实时性和精度的要求。

Method of Monocular Vision Measurement for UAV Pose Based on Characteristic Points

As to the problem of rapid and high-precision positioning of drones in dynamic environment, this paper proposes a method of performing pose and position calculation on the artificial features of UAV with a monocular camera. A number of small LED lamps are placed on the drone as feature points for visual measurement. One of those points is used as the origin to establish the UAV body coordinate system. The multi-scene measurement is used to determine the three-dimensional position of the feature point in the body coordinate system. The image points in the image are then matched to the points in the world. The position and pose of the drone are solved by EpnP algorithm at last. In the experimental part, with the application of three-axis motion platform and two-dimensional turntable, the errors of position solution result and the pose calculation result are respectively measured. The experimental results show that the position solution error is less than 2%, and the pose error is about 8%, with the processing time of merely 2ms, and prove that this algorithm can meet the real-time and accuracy requirements from UAV positioning.