中文摘要：

为实现空间中无法直接获取位姿信息的非合作航天器的相对导航，在充分考虑目标星在空间中做自由翻滚的前提下，提出一种基于立体视觉的超近距离非合作航天器相对导航理论。根据相对位姿动力学模型推导其状态方程，利用立体视觉系统提供观测数据，在此基础上设计扩展卡尔曼滤波器，确定目标卫星相对于追踪星的相对位置，相对速度和角速度，从而优化和引导机械臂捕获目标航天器，并通过实例进行仿真验证。仿真结果表明：该方法对相对位置的估计精度优于0．01m，相对姿态精度优于0．02°，能有效提高超近距离非合作航天器相对导航的精度。

英文摘要：

For realizing relative navigation of non-cooperative spacecrafts which can＇t acquire position and attitude information, considering target star free to tumble at space, proposes a theory that super close relative position and attitude determination between non-cooperative spacecrafts based on stereo vision. According to relative position and attitude dynamics model, deduce state equation, use stereo vision system to provide observation data. Based on this, design extended Kalman filter. Ensure relative position of relative satellite to tracking star, relative speed and angular velocity. Optimize and guide manipulator to capture the target spacecraft, then use example to carry out simulation verification. The simulation results show the stability of designed system, and the estimation accuracy of relative position in this paper is better than 0：01 m, accuracy of relative attitude reaches 0.02°. The filtering algorithms in this paper can effectively improve the accuracy of ultra-close relative navigation for non-cooperative spacecraft.