SVD算法在建筑物点云数据配准中的应用

Application of SVD algorithm in registration of building point clouds

点云数据配准是点云数据处理的一个关键性环节。配准的目的就是为了得到不同视角下点集之间的平移向量与旋转矩阵。SVD算法(Singular Value Decomposition,奇异值分解法)是一种可靠的求解平移向量与旋转矩阵的方法。本文通过对建筑物扫描点云数据进行配准,从点云重叠度、噪声、初始配准状态三个方面讨论SVD算法的精度和时间消耗问题,实验结果可以作为点云数据预处理很好的参照。

The data registration is a critical link in data processing of point clouds. The purpose of registration is to obtain a vector of translation and rotation matrix under different perspectives. The SVD algorithm is a reliable algorithm of solving the translation vector and rotation matrix. In this paper,through registration of building point clouds,the accuracy and time consumption of this method are discussed based on overlap,noise and initial registration status of point clouds. The experimental results can be used as a good reference for preprocessing of point clouds.