基于采样一致性的点到面度量点云配准算法
Point-to-Plane Metric Point Cloud Registration Algorithm with Sampling Consistency
武斌 1刁兴琳 1赵洁 1王姝臻1
作者信息
- 1. 天津城建大学计算机与信息工程学院,天津 300384
- 折叠
摘要
针对目前点云配准算法依赖初始位姿、收敛速度时间慢、精度不够高等问题,提出了一种基于采样一致性算法(SAC-IA)的点到面度量点云配准算法.首先,对原始点云数据进行体素均匀降采样,在一定程度上精简点云数据量,降低计算成本;其次,利用基于法向量夹角阈值的快速点直方图(FPFH)进行特征点提取及特征描述,同时结合采样一致性算法计算出点云的初始变换矩阵;最后,在初始变换的基础上采用点到面度量配准算法完成精细配准.实验结果表明,该算法相较于基于KD-tree加速的ICP算法,显著提升了配准精度,具有较高的配准效率.
Abstract
Point-to-Plane metric point cloud registration algorithm with sampling consistency initial alignment(SAC-IA)is pro-posed to address the problems of the current point cloud alignment algorithm relying on initial poses,low convergence speed time and insufficient accuracy.Firstly,the raw point cloud data are uniformly downsampled using voxels to reduce the dataset size and computational overhead.Secondly,feature points are extracted and characterized using Fast Point Feature Histograms(FPFH)with a normal vector angle threshold.Meanwhile,the initial transformation matrix for the point cloud is then deter-mined via SAC-IA.Finally,Point-to-Plane metric point cloud registration algorithm is used to complete the fine alignment based on the initial alignment.The experimental results show that the algorithm significantly improves the alignment accuracy and has higher alignment efficiency than the ICP algorithm based on KD-tree acceleration.
关键词
点云配准/FPFH快速点直方图/SAC-IA采样一致性/点到面度量Key words
point cloud registration/fast point feature histograms/sample consensus initial alignment/point-to-plane measurement引用本文复制引用
出版年
2024
NETL
NSTL
万方数据