摘要
月球探测器在下降着陆过程中探测到的月面图像灰度均匀,缺少纹理信息,而视觉辅助导航采用传统ORB算法对灰度均匀的月面图像检测出的特征点数量较少,容易出现团簇现象,因此提出一种基于改进ORB算法的图像特征提取与匹配算法。首先对月面图像构建金字塔尺度空间模型,设置自适应阈值并构建掩膜逐一分块检测特征点;然后采用非极大值抑制使特征点分布均匀,消除聚集现象再进行特征匹配;最后结合渐进一致采样算法剔除误匹配,并计算下降序列图像间的位置变换关系。实验结果表明:该方法对于存在大尺度、旋转变化情况下的月面图像,在匹配精度方面比ORB提高10%~13%,在月面着陆视觉导航应用方面具有较好的适用性。
Lunar surface images detected by landers during the descent are of uniform grayscale,lacking texture information.In addition,the number of feature points of lunar surface images with uniform grayscale detected by the visual navigation in terms of the traditional oriented FAST and rotated BRIEF(ORB)algorithm is small,and the phenomenon of cluster is easy to appear.Therefore,an image feature matching algorithm based on improved ORB algorithm is proposed.First,a pyramid scale space model is constructed for the lunar surface images,and the adaptive threshold detection feature points are set.Then,non-maximum value suppression is used to make the feature points uniformly distributed,and the Hamming distance is used for feature matching.Finally,the progressive sample consensus algorithm is used,the matching pair is purified,and the transformation matrix between the descending sequence images is calculated.The experimental results show that the matching accuracy of the proposed method is10%~13%higher than that of ORB for the lunar images with large scale and rotation change,and the method has good applicability in the application of lunar surface landing visual navigation.
作者
胡涛
贺亮
夏永江
HU Tao;HE Liang;XIA Yongjiang(Shanghai Institute of Spaceflight Control Technology,Shanghai 201109,China;Shanghai Key Laboratory of Aerospace Intelligent Control Technology,Shanghai 201109,China;Shanghai Academy of Spaceflight Technology,Shanghai 201109,China)
出处
《上海航天(中英文)》
CSCD
2021年第2期45-51,共7页
Aerospace Shanghai(Chinese&English)
基金
载人航天预先研究项目(060201)。
关键词
月面着陆
特征提取
视觉导航
改进ORB
自适应阈值
非极大值抑制
lunar surface landing
feature extraction
visual navigation
improved oriented FAST and rotated BRIEF(ORB)algorithm
adaptive threshold
non-maximum suppression