一种基于特征点的地面激光点云全局配准方法

A Global Registration Method of Terrestrial Laser Point Cloud Based on Feature Points

针对当前地面激光扫描(TLS)点云配准自动化程度低且耗时的问题,本文提出一种基于特征点和改进FGR(fast global registration)算法的TLS点云全局配准方法。该方法一共分为三步:首先对点云进行粗差剔除和降采样;然后提取Do G(Difference-of-Gaussian)特征点和进行FPFH(fast point feature histogram)描述,进而进行双向一致性匹配;最后使用FGR算法进行优化获得点云之间初始参数,结合标准ICP算法实现TLS点云的高精度配准。利用7站地面激光点云数据进行实验,结果表明本方法可以在保证配准精度的前提下获得较高的配准效率。

Aiming at the problem of low degree of automatic and time consuming to align pair-wise of TLS point cloud,a global registration method based on feature point and FGR（ fast global registration） algorithm is proposed. The method consists of three steps： at first,the outlier is removed and down-sampling is carried out; After that,the Do G（ Difference-of-Gaussian） feature is extracted and the FPFH（ fast point feature histogram） is described,then matching based on FPFH is performed; Finally,FGR algorithm is used to obtain the initial parameters and standard ICP algorithm is used to optimize the transformation parameters on original point cloud. Seven stations of TLS data are employed to perform the experiment,the results reveal that the method can improve the registration efficiency ensuring the registration accuracy.