虚实配准是增强现实的关键技术。因为有运算量小实时性强的特点，利用平面模板进行虚实配准被广泛使用。本文算法利用平面模板进行虚实配准中的单应性映射关系，建立从平面模板坐标系到摄像机坐标系3维空间坐标转换和从摄像机坐标系到投影平面的3维-2维空间坐标转换，既减小了运算量，又解决了人工定标的问题。在实现虚实配准的基础上，进一步探讨了配准的优化问题，对关键的单应性矩阵进行优化筛选，提高虚实配准的强劲性，实现部分标志点被遮挡情况下的虚实配准。实验结果证明，本文算法成功实现了虚实配准，并在抗遮挡性上有强劲的表现。

Registration of virtual and real objects is a crucial technology for augmented reality. Because pattern-based algorithm has less operation cost, the method has been used widely in augmented reality. Homography is a mapping relation existing in the pattern-based registration. It has been employed to build the 3D transformation from pattern coordinate to camera coordinate and the 3D-2D transformation from camera coordinate to project coordinate in this paper. As a simple proportional relation, homography cuts back the operation cost. Moreover, it solves the problem of manual calibration. After elaborating the registration course, registration optimization has been further discussed in the paper. The best homography will be selected from several computed results in order to enhance the robustness of registration. The optimization ensures that registration succeeds with partial mark points occluded.