中文摘要：

对两幅不同视角下的点云采用面表法管理，而后对两幅不同视角下的点云采用曲率特征的ICP改进算法进行配准处理.具体处理方法为先根据获得的两幅海量点云数据进行面表法处理，然后根据获得的面表计算每幅点云中的近似曲率，对其中一幅点云的曲率按大小分类，并且每类进行随机采样.根据随机采样的结果按照曲率第一、距离第二的原则在另一幅点云中查找对应点，最后求出第一幅点云的旋转和平移矩阵从而完成配准过程.实验结果表明：在相同海量点云下，面表法的生成时间比八叉树法生成时间平均少406 ms，查找时间平均少53.5 ms；并且曲率特征的ICP算法可以迭代25次收敛到0.241625 mm的精度，基本满足多视角点云配准的精度高、计算速度快等要求.

英文摘要：

Two point clouds in two different perspectives is managed by the method of surface and then the ICP algorithm is adopted based on curvature characteristic to improve the registration process. The surface method for processing the acquired point clouds from two different perspectives is used firstly. Then the curvature of two point clouds based on the surface method is calculated,and classify one of two point clouds by the curvature size and sample randomly from different classes. According to the result of random sample, to find the corresponding points which is in another point cloud in accordance with the principle of the curvature first, distance second. At last, calculate rotation matrix and translation matrix of the first point cloud to complete the registration process. Experimental results show that under the condition of the same mass point cloud, surface table generation time is less than the octree generation time on average 406 ms while the finding time is less than average 107 ms. What′s more, the ICP algorithm for curvature features based on curvature features achieves the precision of 0.241 625 mm. It can be basically satisfied the requirements of the multi-view point cloud of high registration precision , fast computing speed, etc.