首页 >

翻滚非合作卫星稠密重建与位姿参数估计方法-AFC2026,正刊

牟金震,张赫,韩飞   

  1. 上海航天控制技术研究所
  • 收稿日期:2026-05-26 修回日期:2026-09-23 出版日期:2026-09-28 发布日期:2026-09-28
  • 通讯作者: 牟金震
  • 基金资助:
    国家自然科学基金

Dense Reconstruction and Pose Parameter Estimation of Tumbling Un-cooperative Satellites

  • Received:2026-05-26 Revised:2026-09-23 Online:2026-09-28 Published:2026-09-28

摘要: 针对完全未知的翻滚非合作卫星稠密重建与位姿参数估计问题,提出一种基于面元重建与变形图优化的位姿参数估计方法。首先,利用迭代最近点算法初始化位姿并使用面元表示重建模型。针对空间光照特性,利用特征配准时图像信息重要程度加权实现位姿联合跟踪。根据失效卫星的翻滚特性,利用随机蕨算法对当前帧编码,并与编码数据库比对,判断卫星是否存闭环。若存在闭环,应用变形图优化模型参数,消除累计误差。基于关键帧序列绕进动轴旋转的特点,利用关键帧轨迹拟合圆轨迹并确定进动轴的方向和位置。实验结果表明,所提方法准确地辨识出翻滚卫星的运动参数且能够实现连续稳定的长时间测量,满足工程需求。

关键词: 非合作翻滚卫星, 稠密重建, 位姿跟踪, 变形图优化, 翻滚参数估计

Abstract: To solve the problem of density reconstruction and pose parameter estimation of a completely unknown tumbling non-cooperative satellite, a surfel reconstruction and a deformation graph optimization are proposed. Firstly, the Iterative Closest Point (ICP) algorithm is used to initialize the pose and the surfel reconstruction is used to reconstruct the satellite. According to the characteristics of space illumination, a weighted method of feature matching on image information im-portance degree in adaptive adjustment is proposed to track pose. According to the tumbling characteristics of the inactive satellites, the random fern algorithm is used to encode the current frame and compare with the coding database to judge whether the satellite has a closed loop. If closed loop exists, the parameters of the reconstructed model are optimized by using deformation map to eliminate accumulated error. Because the key-frames rotate around the precession axis, the circle fitting method is used determine the direction and position of the precession axis. The simulation results show that the proposed method can accurately identify the motion parameters of the tumbling satellite and achieves long time stability measurement, which meets the engineering requirements.

Key words: On-cooperative tumbling satellite, Dense reconstruction, Pose tracking, Deformation map optimization, Tum-bling parameter estimation