基于松弛极线约束的月面复杂仿射变换图像匹配方法  

An epipolar relaxation constrained matching algorithm of largeaffined images for lunar rover with large span distance in a single movement

作　　者：刘传凯 王沼翔[1] 雷俊雄 张作宇 樊宽刚 张济韬 王晓雪 潘海朗 刘建国[5] LIU Chuankai;WANG Zhaoxiang;LEI Junxiong;ZHANG Zuoyu;FAN Kuangang;ZHANG Jitao;WANG Xiaoxue;PAN Hailang;LIU Jianguo(School of Electrical Engineering and Automation,Jiangxi University of Science and Technology,Ganzhou 341000,China;Beijing Aerospace Control Center,Beijing 100194,China;The Key Laboratory of Science and Technology on Aerospace Flight Dynamics,Beijing 100194,China;School of Electronic and Optical Engineering,Nanjing University of Science and Technology,Nanjing 210094,China;Department of Earth Science and Engineering,Imperial College London,London SW72AZ,United Kingdom)

机构地区：[1]江西理工大学电气工程与自动化学院,赣州341000 [2]北京航天飞行控制中心,北京100194 [3]航天飞行动力学技术重点实验室,北京100194 [4]南京理工大学电子工程与光电技术学院,南京210094 [5]帝国理工学院地球科学工程系,伦敦SW72AZ

出　　处：《航空学报》2024年第2期223-236,共14页Acta Aeronautica et Astronautica Sinica

基　　金：国家自然科学基金(61972020,62003025);实验室基金(6142210200307,XTB6142210210302)。

摘　　要：月球车在月面巡视探测过程中采用大间距行进模式、以倾斜视角拍摄月面,导致相邻站点间所拍摄的月面图像重叠区域较小,并且在尺度、旋转上具有较大的差异;同时由于站点间图像仿射变换复杂、纹理相似度高且光照条件不一致,仅利用图像局部表观特征匹配极易发生错误。针对上述问题,提出了基于双向松弛极线约束的图像匹配算法,首先,将惯导测量的站点间近似相对位姿引入到成像光束的投影变换中,计算一个站点图像特征点在另一站点图像中对应的极线方程;其次,根据站点误差范围近似估计前后站图像中双向极线的松弛约束范围,构建限定在松弛约束范围内的匹配候选点集合;再次,利用FLANN算法进行特征匹配,选取同时满足双向松弛极线约束且相似度高的特征点对;最后,使用RANSAC算法进一步剔除错误匹配点,得到最终配对点集合。实验表明:上述算法在处理站点间大尺度、旋转变换月面图像匹配问题时,能够尽可能多地保留正确匹配点,极大提升匹配正确率和剔除无效匹配点,相比ASIFT等算法效果提升明显。In the process of lunar surface exploration,Yutu-2 adopts a widely spaced travel mode and takes photos of the lunar surface from a tilted perspective,resulting in a small overlap area of the lunar surface images taken between neighboring sites,and large differences in scale and rotation.In addition,due to the complex image affine transforma⁃tion between sites,high texture similarity and inconsistent illumination conditions,it is easy to make errors by matching only local apparent features of images.To address the above problems,this paper proposes an image matching algo⁃rithm based on the bi-directional relaxation polar line constraint.Firstly,the approximate relative poses between sta⁃tions measured by inertial guidance are introduced into the projection transform of imaging beam,and the polar line equation corresponding to the feature points of one station image in another station image is calculated.Secondly,the bi-directional polar line relaxation constraint range in the front and back station images is estimated based on the sta⁃tion error range,and a set of matching candidates limited within the relaxation constraint range is constructed.Then,the FLANN algorithm is used to select the feature pairs that satisfy the bi-directional relaxation constraint and have high similarity.Finally,the RANSAC algorithm is used to further reject the wrong matching points and obtain the final set of matching points.The experiments show that when dealing with the problem of inter-site large scale and rotationally transformed lunar surface image matching,the proposed algorithm can retain as many correct matching points as pos⁃sible,greatly improve the correct matching rate and eliminate invalid matching points,and has a significant improve⁃ment in effect compared with ASIFT and other algorithms.

关 键 词：月面探测 月面图像 仿射变换 双向松弛极线约束 图像配准 

分 类 号：V448.2[航空宇航科学与技术—飞行器设计]