点云数据的自动配准算法研究

Research on Automatic Registration Algorithm of Point Cloud

针对传统ICP(Iterative Closest Points)配准算法计算量大、收敛速度慢且要求待配准的两片点云数据重合程度较高的问题提出了一种改进方法:首先基于均匀采样法精简点云数据;其次采用Kd-Tree算法查找最近点并基于距离阈值剔除错误匹配点;接着优化目标误差函数,计算点到切平面的距离;最后采用多角度的全局配准方法将两片重合程度最小的点云较好地配准在一起。通过对比实验,验证了本文的改进型ICP算法在运行时间和配准精度上都对传统的ICP算法做出了较大改进,取得了较好的配准效果。

This paper proposes an improved method to solve the problem of traditional ICP(Iterative Closest Points)registration algorithm that has a large amount of calculation,slow convergence speed and high degree of coincidence of the point clouds to be registered:firstly,the point cloud is simplified based on the uniform sampling method;secondly,the Kd-Tree algorithm is used to find the closest point and the wrong matching point is eliminated based on the distance threshold;then,the distance from the point to the tangent plane is calculated to optimize the objective error function;finally,the multi-angle global registration method is used to better register the two point clouds with the smallest degree of overlap.It is verified by experimental comparison that the improved ICP algorithm used in this paper has made great improvements to the traditional ICP algorithm in terms of running time and registration accuracy,and has achieved better registration results.