若要实现空间失效卫星、空间碎片等非合作目标, 尤其是具有自旋运动特性的目标的在轨服务或者离轨清除, 需要精确测量追踪航天器与目标之间的相对姿态。当目标旋转一周后, 对重访区域的闭环检测与位姿优化是减小累积误差, 提高测量精度的重要保证。首先, 本文介绍了视觉词袋库的建立和基于视觉词袋的非合作目标闭环检测策略。然后, 基于相似性变换, 对图像序列关键帧集和当前帧进行了联合位姿图优化, 实现了对刚体变换矩阵的校正。最后对不同运动角速度下的不同目标, 采用不同相机进行地面模拟测试实验, 验证了方法的有效性和可靠性。实验结果表明: 对于以12(°)/s角速度运动的非合作目标, 当测量稳定后, 绝对角度误差约为1(°), 相对角度误差约为1%, 平均角速度误差约为0.12(°)/s。可以满足非合作目标相对姿态测量的任务需求。

To achieve on-orbit service or off-orbit elimination of non-cooperation targets such as failed satellites and space debris, especially the rotating targets, it is required to carry out accurate measurement of relative pose between the tracking spacecraft and the target. When the target rotates for a round, closed-loop detection and pose optimization in the revisit region serve as a significant guarantee for lowering the accumulative error and improving the measurement accuracy. First, establishment of the bag of visual words and strategy for closed-loop detection of non-cooperation targets according to such bag of visual words were introduced in this paper; combined pose optimization was carried out in key frames and current frame of the image sequence based on the similarity transformation, thus achieving correction of the rigid transformation matrix; finally, ground-based simulation test was conducted in different targets under different angular speed by using different cameras to verify effectiveness and reliability of the proposed method. The results indicate that for non-cooperation targets with an angular speed of 12(°)/s, its absolute angle error was about 1(°), with a relative angle error of about 1% and an average angle error of about 0.12(°)/s when the measurement stabilizes. Hence, proposed method can satisfy the mission requirements for relative pose measurement of non-cooperation targets.