非合作慢旋目标即时状态位姿确定

对失效卫星等非合作慢旋目标进行在轨服务，需要精确测量追踪航天器与目标之间的姿态信息。因此，如何在复杂的光照条件下快速、准确地对非合作慢旋目标进行即时状态位姿确定具有一定的挑战性。应用ORB-SLAM技术，首先定位关键帧，并估计姿态。然后用当前帧的特征点与地图点对应的特征点进行匹配。最后，将完成匹配的特征点通过重投影确定其在地图中的位置，如果出现跟踪丢失，则根据已有的地图点估计姿态。实验结果表明：在复杂的光照条件下，分别对以10(°)/s角速度运动和以3(°)/s角速度运动的非合作目标进行测量，当测量稳定后，平均角速度误差约为0.1(°)/s和0.02(°)/s，可以满足工程上空间非合作目标相对姿态测量的精度要求。

Real-time Pose Estimation for Slow Rotating Non-cooperative Targets

To achieve on-orbit service of slow rotating non-cooperative malfunction satellites, it is required to acquire the accurate measurement of the attitude information. Therefore, it is a challenging to estimate pose information of slow rotating non-cooperative targets rapidly and accurately under complex illumination conditions. In this paper, an ORB-SLAM based method is used. First, locate the key-frames and estimate the pose. Then the feature points of the current frame are matched with the corresponding feature points of the map. Finally, the position of the matching feature points in the map is determined by re-projection. If tracking loss occurs, the pose is estimated according to the existing map points. Experimental results show that the stable measurement errors of the mean angular velocity are about 0.1(°)/s and 0.02(°)/s for a target rotating at the speed of 10(°)/s and the speed of 3(°)/s respectively under complex illumination conditions. It can meet the system requirements of pose estimation for the non-cooperative targets.