Abstract: Point cloud registration is one of the key technologies of simultaneous localization and mapping (SLAM) based on lidar. The ultimate goal of point cloud registration is to solve the transformation matrix of different posture point clouds in the same coordinate system, and use this matrix to achieve accurate registration of multiple scan point clouds, so as to obtain the transformation of vehicle body position and posture. This article first introduces the idea of point cloud registration, summarizes and explains the development status of iterative closest point (ICP) algorithm and normal distribution transformation (NDT) algorithm, and then mainly introduces two.

Key words: Point cloud registration; Iterative Closest Point algorithm; Normal Distribution Transformation algorithm

摘要： 点云配准是基于激光雷达的同时定位与建图（SLAM）的关键技术之一。点云配准最终目的是求解出相同 坐标系下不同姿态点云的变换矩阵，利用该矩阵实现多次扫描点云的精确配准，从而获得车身位置和姿态的变换。 文章首先介绍点云配准的思想，对迭代最近点（ICP）算法和正态分布变换（NDT）算法的发展现状概括并进行说 明，然后主要介绍 ICP 和 NDT 两种经典点云配准算法，最后进行总结。

关键词: 点云配准；迭代最近点（ICP）算法；正态分布变换（NDT）算法