采用单目视觉的非合作目标星状态估计

针对微小卫星逼近观测未知的空间翻滚非合作目标星任务，提出一种基于单目视觉的目标星相对状态估计方法。在建立追踪星/目标星相对运动模型的基础上，以单目相机识别并测量获得的目标星固有特征的像素位置为观测输入，通过扩展卡尔曼滤波算法实现对目标星相对位置、相对速度、相对姿态、角速度、惯量比和特征位置等状态的估计。仿真结果表明,该方法能够很好地实现对未知非合作目标星的相对状态估计，姿态估计误差小于2°，位置估计误差小于0.1 m，特征点位置平均估计误差小于0.04 m。

State Estimation of a Non Cooperative Target Satellite Using Monocular Vision

Considering the mission of a microsatellite approaching an unknown and non-cooperative target satellite, a method for relative state estimation of the target satellite using a monocular camera is proposed. Firstly, the relative kinematics and dynamics models between the target and the tracking satellite are established. Then, the pixel location of the feature points on the target is used as the filter input, and the relative states including relative position, relative velocity, relative attitude, angular velocity, inertia ratio and feature position of the target are estimated. Finally, the simulation results demonstrate the validation of the proposed method with the attitude estimation error less than 2°, the position estimation error less than 0.1 m and the average position estimation error of feature points less than 0.04 m.