期刊文献+

基于主动特征选择的非合作航天器鲁棒视觉导航方法研究

Robust Method Study of Active Feature Selection for Non-Cooperative Spacecraft Vision-Based Navigation
下载PDF
导出
摘要 面向非合作目标航天器近距离操作任务,针对采用自然特征的单目视觉相对位姿参数确定过程中特征提取与匹配导致的粗大误差增加导致结果不准确甚至错误,以及特征数量多增大计算量等问题,提出一种融合随机采样一致性(RANSAC)算法和主动特征选择的鲁棒视觉导航方法。用RANSAC算法剔除有粗大误差的特征点,给出了基于RANSAC的特征点选择步骤;根据不同特征点组合所计算的克拉美罗(CRLB)不同,用参数化CRLB下限选择对位姿确定精度有显著影响的点以减少参与计算的特征数量,给出了基于CRLB的特征点选择流程。仿真结果表明:综合RANSAC和CRLB的特征点选择方法可显著减少特征点数量,提高了位姿解算精度。 To solve the problem that the error of feature points extracting or matching and the number of feature points would lead to inaccurate results or the wrong results and huge amount of calculation on the relative position and attitude parameter determination during non-cooperative target spacecraft proximity operations,a robust method for vision navigation fusing the random sample consensus(RANSAC)algorithm and an active feature selection method was put forward in this paper.First the gross error was eliminated by RANSAC algorithm.The selection steps for feature points were given based on RANSAC algorithm.Then the different points,which had significant impact on determining precision based on Cramér-Rao lower bound(CRLB),were selected to reduce the number of feature involved in the calculation according to the different CRLB calculated from various feature points sets.The features selection flowchart was given based on CRLB.The simulation results showed the selected points by combining of RANSAC and CRLB could be reduced and the precision of position and attitude had been improved.
出处 《上海航天》 2016年第6期136-141,共6页 Aerospace Shanghai
基金 上海航天科技创新基金资助(SAST201444)
关键词 非合作目标 视觉导航 特征点选择 RANSAC CRLB 特征点数 鲁棒性 位姿精度 Non-cooperative spacecraft Visual navigation Feature point selection Random sample consensus(RANSAC) Cramér-Rao lower bound(CRLB) Feature point number Robust Precision of position and attitude
  • 相关文献

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部