中文摘要：

点特征匹配在机器视觉、图像配准等领域中有着重要的应用．针对空间存在较大仿射几何差异的图像中的点特征匹配问题，提出了一种利用马氏距离仿射不变性进行约束的松弛匹配算法，并将该算法应用于遥感图像配准中．实验结果表明，算法可以很好的完成点特征匹配，匹配点对数量充足且具备很高的正确率，从而可以保证图像配准的精度．

英文摘要：

Point pattern matching is a common problem in machine vison and image registration. To match points from images with relatively great affine geometric distortion, a rdaxation algorithm restricted by the invariance of Mahalanobis distance in affine geometric distortion is proposed and used in the registration of remote sensing image. Experimental result shows that sufficient correct matching-points are accquired by the algorithm and can be applied to registration of images with relatively great affine geometric distortion.