中文摘要：

针对传统的迭代最近点算法（ICP）用于多测站点云数据配准时计算效率低的问题,该文提出了一种基于特征点的ICP改进算法,该方法利用相邻两测站数据进行配准的实现.首先采用体素化格网方法对两点云数据集进行精简处理,并计算精简处理后每一点的法向量;然后利用kd-tree最近邻查询搜索特征点之间的对应关系;并通过估计出的最优变换矩阵更新至全局变换,以提高配准精度.实验结果表明,改进的ICP算法在地铁隧道点云数据配准中的效率高于其他的配准方法,为隧道变形监测工作的进行提供保证.

英文摘要：

For the computational efficiency of traditional Iterative Closest Point （ICP） algorithm used in the registration of the two adjacent stations was low, this paper proposed an improved ICP algorithm based on feature points, which used the two adjacent station data to realize the registration： first voxel grid method was used to streamline the two point cloud data sets, the normal vector of the points was cal- culated after streamlined, and then the corresponding relationship of feature points was searched by using kd-tree nearest neighbor query and optimal transformation matrix was estimated to update the global transformation. The experiment result showed that the efficiency and precision of the improved ICP algorithm was higher than other algorithms in the registration of tunnel point clouds data, which could provide guarantee for the tunnel deformation monitoring.