中文摘要：

三维点云配准是三维面型反求中的关键和难点。提出了一种特征点三维点云配准技术。通过引入被测物的特征点，分析了坐标变换矩阵的求解方法，利用最小二乘算法求出初始变换矩阵，得到粗略配准结果。然后采用K—D树来加速搜索最近点，用均方距离作为误差度量准则实现了改进后的最近点新的迭代算法，进一步得到了很好的精确配准效果，并给出了实现这种算法的程序设计思想。利用鞋楦配准实验证明了此方法的有效性和可靠性。

英文摘要：

The registration of 3-D point clouds is the key and difficult problem in 3-D surface reverse. A registration method of 3-D point clouds based feature points is put forward. The method of solving the coordinate transformation matrix is analyzed. The initial transformation matrix is calculated with Least -squares algorithm by introducing feature points. The improved ICP algorithm leads to a more accurate effect after K-D tree on finding closet points and a better error criterion. An idea concerning the pseudo-code for the program design is also presented. In addition, the method is carried out by last registration test and proved to have a perfect effectiveness and reliability.