摘要
为提高月球探测器自主软着陆的落点精度,利用绕月飞行器和着陆探测器下落图像,提出一种基于尺度信息的多模板递阶景象匹配方法.在月面图像尺度空间上快速提取FAST角点,并确定特征点精确尺度、位置和方向.在特征点邻域建立图像块采样模式,通过图像块像素比较形成二进制串特征描述子,利用海明距离(Hammingdistance)进行特征匹配.实验结果表明,目标在尺度缩放、旋转和光照变化等极端条件下,该算法能够实时完成月面着陆目标区域的准确识别和稳定跟踪,实现月球探测器的远距离、高精度自主导引着陆.
In order to improve the accuracy of autonomous soft landing on the lunar surface, a multi-model hierarchical scene matching method based on scale theory is proposed, using the images taken by the lunar orbiter and lander as the information sources. The FAST keypoint extracting technique is used in scale space of the lunar surface image, and then the exact scale, location and orientation of the keypoints are decided. A patch pair sampling pattern around the keypointis is introduced to build the binary descriptor with comparisons of patch pairs. Matching two descriptors is a simple computation of their Hamming distance. Experiments indicate that, with its robustness to scale changes, rotation, motion blur and various illumination conditions, the proposed method could well fulfill the real-time landing target recognition and tracking needs of automatic and precise soft landing on the lunar surface.
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2013年第2期172-177,共6页
Transactions of Beijing Institute of Technology
基金
国家部委预研项目(51309020402)
作者简介
郑智辉（1986-），男，博士生，E—mail：zhengzhihui009@qq．com
汪渤（1963-），男，教授，博士生导师，E-mail：wangbo@bitedu．cn．
