摘要
针对传统迭代最近点算法(ICP)对点云初始位置要求高、收敛速度慢和易陷入局部最优的问题,本文提出了一种基于特征点采样一致性算法改进ICP算法的点云配准方法。首先使用体素网格法采样,通过法向量邻域夹角特性提取特征点并建立快速点特征直方图(FPFH)进行特征描述;然后使用采样一致性算法(SAC-IA)粗配准计算出点云的初始坐标变换,进而使点云获得较好的初始位置;最后通过K维树近邻搜索改进ICP算法,完成点云精确配准。实验结果表明,所提出的方法能够提供良好的初始位置,提高传统ICP算法点云的配准精度和收敛速度。
Aiming at the problems that the traditional iterative closest point algorithm(ICP)requires accurate initial position of the point cloud,slow convergence speed and easy to fall into local optimum,this paper proposed a point cloud registration method based on feature point the Sample Consensus Initial Alignment(SAC-IA)and improved ICP algorithm.First down-sampling using the voxel grid method,feature points were extracted by the angle characteristic of the normal vector neighborhood and a fast point feature histogram(FPFH)was established for feature description;Then using the SAC-IA registration to calculate the initial coordinate transformation of the point cloud,so that the point cloud can be obtained better initial position;Finally,the ICP algorithm was improved by K-dimensional tree nearest neighbor search to complete the point cloud accurate registration.According to the experimental results,this method could provide a good initial position,improved the registration accuracy and convergence speed of the traditional ICP algorithm.
作者
宋成航
李晋儒
刘冠杰
SONG Chenghang;LI Jinru;LIU Guanjie(College of Geomatics,Shandong University of Science and Technology,Qingdao Shandong 266590,China)
出处
《北京测绘》
2021年第3期317-322,共6页
Beijing Surveying and Mapping
