一种适合降落序列影像的快速匹配方法

A Fast Matching Method for Descent Imagery

针对嫦娥三号降落序列影像特征匹配耗时较多的问题,提出了一种适合降落影像的快速匹配方法。首先对待匹配的降落影像进行尺度变换,将二者进行统一,然后采用9×9大小的窗口进行基于Moravec方法的特征点提取。其次,基于降落过程几何模型约束,以相关系数为匹配测度,实现降落影像间的匹配。采用真实嫦娥三号降落影像开展了匹配实验。实验结果表明,特征点窗口大小为9×9,相关系数阈值为0.90时,提出匹配方法的性能最佳;此外,与经典SIFT匹配相比,提出的方法可以平均减少59%左右的匹配耗时,且误匹配率更小。研究成果对我国后续深空探测任务中降落影像的匹配工作具有重要的参考价值。

Aiming at the time-consuming problem of image matching in Chang E III descent images,a fast matching method suitable for landing images is proposed.First,scale transformation is applied to match the descent images,in order to unify the scale of two images.Then,a 9×9 size window is used to extract feature points based on Moravec method.Finally,the matching between the descent images is realized based on the geometric model constraint of landing process under the correlation coefficient.The matching experiment was carried out with the real Chang E III descent images.The experimental results show that the matching method performs best when the size of the feature point window is 9×9 and the correlation coefficient threshold is 0.90.In addition,compared with the classic SIFT matching,the proposed method can reduce the average time of the match by about 59%,and the mismatch rate is smaller.The research results have important reference value for the follow-up deep space exploration task.