汽车钣金件三维点云位姿估计技术研究
Research on 3D Point Cloud Pose Estimation Technology for Automotive Sheet Metal Parts
王海亮 1郑双 1于军波1
作者信息
- 1. 威海广泰空港设备股份有限公司,山东 威海 264200
- 折叠
摘要
针对传统汽车钣金件三维点云位姿估计问题,本研究提出了采用随机采样一致性配准算法(SAC-IA算法)进行点云粗配准,获取到较为精准的初始姿态后,采用多维二叉查找树(K-D tree)对迭代最近点算法(IPC算法)进行改进,实现点云精准配准,获取到精度更高的点云配准方案,并用于汽车钣金件三维点云位姿估计.实验结果表明,本研究方法相比其他配准方法的综合配准性能更佳,实际应用的位置误差小于 4mm,角度误差小于 4.5°,计算时间小于 6s,可以满足工程要求.
Abstract
In response to the problem of traditional 3D point cloud pose estimation for automotive sheet metal parts,this study proposes the use of the Random Sampling Consistency Registration Algorithm(SAC-IA algo-rithm)for point cloud coarse registration.After obtaining a more accurate initial pose,a Multidimensional Bina-ry Search Tree(K-D tree)is used to improve the Iterative Nearest Point Algorithm(IPC algorithm),achieving precise point cloud registration and obtaining a more accurate point cloud registration scheme,which is used for 3D point cloud pose estimation of automotive sheet metal parts.The experimental results show that this research method has better comprehensive registration performance compared to other registration methods,with a practi-cal application position error of less than 4mm,an angle error of less than 4.5°,and a calculation time of less than 6 seconds,which can meet engineering requirements.
关键词
汽车钣金件/三维点云/位姿估计技术Key words
Automotive sheet metal parts/3D point cloud/Pose estimation technology引用本文复制引用
出版年
2024
NETL
NSTL
万方数据