非合作航天器间相对位姿的单目视觉确定算法

针对非合作航天器接近操作过程中200m以内的相对位姿参数确定问题，该文充分利用目标航天器已知的结构模型信息，推导了利用单个光学相机实现非合作航天器相对位姿参数测量的迭代算法。与传统的交会对接视觉系统不同，该视觉系统不需要在目标航天器上安装光学特征，而是充分利用了目标航天器的自身结构特征，因此更适用于一般航天器间的相对位姿参数测量。为解决相对位姿单目视觉确定的非线性问题，该文选择了求解非线性问题性能优良的Levenberg—Marquardt算法。最后对该文算法进行数学仿真，仿真结果表明算法的有效性和可靠性，适合航天器在轨实时应用。

Monocular Vision-based Relative Pose Parameters Determination for Non-cooperative Spacecrafts

Using monocular vision system, and the knowledge of the target spacecraft, an iterative method for relative pose parameters is found during the proximity operations of non-cooperative spacecrafts from a distance between 1 and 200 meters. Unlike previous spacecraft-based computer vision techniques, the proposed method does not require specialized markers to be present on the target. Instead, a more general system is proposed, which makes use of the target' s own structure in estimating pose. To deal with the nonlinearity of the monocular vision-based relative pose parameters determination, the Levenberg-Marquardt algorithm is used. The solution is validated by using a mathematical simulation, and the simulation results show that the iterative method is effective and reliable, and the proposed algorithm is more suitable in on-board real-time computation.