论文部分内容阅读
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。
In view of the short-range operation task of non-cooperative target spacecraft, the gross errors caused by the feature extraction and matching in the determination of the relative pose parameters of monocular vision using natural features result in inaccuracies or even errors in the results and increase in the number of features This paper proposes a robust visual navigation method based on RANSAC algorithm and active feature selection. RANSAC algorithm is used to remove feature points with gross error, and the feature points selection procedure based on RANSAC is given. According to CRLB calculated by different combination of feature points, the accuracy of pose determination is determined by the lower limit of parametric CRLB Significantly affecting the point to reduce the number of features involved in the calculation, the CRLB based feature point selection process is given. The simulation results show that the method of feature point selection based on RANSAC and CRLB can significantly reduce the number of feature points and improve the accuracy of pose interpretation.