小天体近距离视觉导航的陆标鲁棒匹配方法

Robust Landmark Matching Method for Visual Navigation Near Small Bodies

针对目前陆标匹配方法难以应对小天体附近极端观测条件的问题,提出了一种陆标鲁棒匹配方法。对通过立体光度测量(Stereo-Photo Clinometry,SPC)法生成的地形陆标的匹配误差进行了理论分析,探讨了陆标的位置误差以及相机的位姿估计误差对匹配结果的影响;基于误差分析对陆标中的表面点进行了优化选取,并提出了加权归一化互相关(Weighted Normalized Cross-Correlation,WNCC)方法,以提高匹配精度、鲁棒性和计算效率;利用高保真度的合成图像序列对比分析了WNCC算法与目前小天体探测任务中广泛使用的归一化互相关(Normalized Cross-Correlation,NCC)算法在极端的尺度、视角和光照变化等条件下的匹配性能。数值统计结果表明,WNCC算法能够高效、鲁棒地为导航系统提供精确的陆标匹配信息。

A robust and efficient landmark matching algorithm was proposed in this paper to deal with the extreme environment near the target asteroid. First,the matching error of the landmark generated by the SPC(StereoPhotoClinometry)technology was analyzed,and the influence of landmark position error and cameras’ pose error on the matching results was discussed. Then,based on the error analysis,the optimal landmark points were selected,and a weighted normalized crosscorrelation(WNCC) algorithm was proposed to obtain accurate matching results robustly and efficiently. Finally,high-fidelity synthetic image sequences were generated to compare the performance of WNCC and the widely used NCC(Normalized CrossCorrelation) algorithm in the previous asteroid missions under a wide range of image scales,viewing geometries, and lighting conditions. The numerical results demonstrate the advance of the proposed method in terms of efficiency,robustness,and accuracy.