为准确获取无人机空中加油对接阶段的相对位姿信息，提出了一种机器视觉辅助的INS/GPS/MV 组合导航方案，研究了机器视觉图像的特征点提取与匹配算法. 引入相对惯导误差，建立了全局滤波器的增广状态方程，并根据杆臂矢量推导了GPS和机器视觉量测方程. 设计了基于联邦滤波器结构的全局多速率扩展卡尔曼滤波算法，以融合多速率传感器信息，并与标准EKF算法进行对比. 仿真结果表明，所提出的算法能有效融合惯导、GPS和机器视觉的测量信息，使导航参数的精度和输出带宽均满足导航系统的设计要求，有利于改善无人机的飞行品质，放宽对飞控系统的性能要求.

To precisely obtain the relative pose of UAV during aerial refueling docking, a machine vision-aided INS/GPS/MV integrated navigation scheme is proposed. Feature extraction and match algorithm of machine vision image are studied. By introducing relative inertial errors, an extended state model of global filter is designed. GPS and machine vision measurement models are established using level-arm vectors. A global multirate extended Kalman filter based on federal framework is designed to realize multirate mutisensor data fusion. Comparison is made between the proposed algorithm and the standard EKF algorithm. Simulation shows that the proposed algorithm can effectively fuse INS/GPS/MV data. The navigation parameter precision and output bandwidth satisfy the requirements of UAV aerial refueling. It can improve UAV qualities and loosen requirements of flight control systems.