中文摘要：

针对月球卫星的景象匹配定位问题，为“嫦娥二号”卫星设计了相应的匹配模式，并构造约束模型进行了匹配矩阵的修正，实现了“嫦娥二号”线阵CCD观测图和美国LRO卫星局部基准图的子图像块大规模匹配。实验中，首先通过构建匹配矩阵和测试库验证了单对图像归一化互信息匹配的可行性和局限性，然后通过约束模型修正，显著性水平取0．1时，使匹配成功率从60．7％提高到89．0％，并且测试库中89．0％的正确匹配对被匹配矩阵所保留。实验结果表明，所提方法对“嫦娥二号”观测图的高效、高成功率匹配可行，能够为进一步的月球卫星位置视觉解算提供可靠的匹配数据。

英文摘要：

To solve the lunar satellite scene matching aided location problem, a CCD line array image matching mode is designed based on the actual constraints from Chang＇ e II lunar satellite in this paper. Firstly, sub image blocks segmented from Chang＇ e Ⅱ CCD line array images and the corresponding local base maps made by the WAC global mosaics from the lunar reconnaissance obiter are matched under a normalized mutual information algorithm, so a basic matching matrix to estimate the effectiveness of matching is built. Then the matching matrix is corrected by a constraint model, and thus matching of Chang＇ e II satellite line array image in a wide range is achieved. In experiments,by constructing a matc- hing matrix and a match-testing library, the feasibility and limitations of normalized mutual information based matching on a single sub-block pair are verified, and then through amendment by the constraint model,the matching success rate increases from 60.7% to 89 （under the significance level of 0.1）, while 89.0% of the whole correct pairs in the testing library are retained. Experimental results show that the proposed matching mode is efficient for Chang＇ e Ⅱ CCD line array images, and can serve for further lunar satellite vision based positioning.