融合BRIEF与ICP点云配准的零部件姿态估计方法
Part Pose Estimation Method Based on Fusion of BRIEF and ICP Point Cloud Registration
田华平 1陈鹏飞 2杨涛 1何培垒 2巩鑫2
作者信息
- 1. 西南科技大学信息工程学院 绵阳 621010;特殊环境机器人技术四川重点实验室 绵阳 621010
- 2. 中国航发四川燃气涡轮研究院 绵阳 621000
- 折叠
摘要
物体姿态估计在增强现实、自动驾驶和机器人操作等领域具有重要意义.在机器人抓取与辅助装配过程中实时并准确地估计目标姿态极其重要,为提高装配过程中零部件的抓取与监控的实时性与精度,研究了基于多视图特征库估计目标姿态的方法,设计融合BRIEF(Binary Robust Independent Elementary Features)特征匹配与ICP(Iterative Closest Point)点云配准的零部件姿态估计方法对零部件姿态角进行估计测量.以法兰、扳手、电钻等零件工具进行实验,实验结果表明融合BRIEF特征匹配与ICP点云配准的姿态估计方法运行耗时约200ms,而姿态估计的最大误差角度为4.5°,能有效保证零部件姿态估计方法的实时性与精度,可应用于机器人抓取作业、辅助装配监控等.
Abstract
Object pose estimation is of great significance in the fields of augmented reality,autonomous driving and robot oper-ation.In the process of robot grasping and assisted assembly,real-time and accurate estimation of the target pose is extremely impor-tant.In order to improve the real-time and accuracy of the grasping and monitoring of parts during the assembly process,the method of estimating the target pose based on the multi-view feature library is studied.A part pose estimation method combining BRIEF fea-ture matching and ICP point cloud registration is designed to estimate and measure the part pose angle.Experiment with parts and tools such as flanges,wrenches,electric drills,etc.The experimental results show that the attitude estimation method fusing BRIEF feature matching and ICP point cloud registration takes about 200 milliseconds to run,and the maximum error angle of pose estima-tion is 4.5 degrees,which is effective to ensure the real-time performance and accuracy of the part pose estimation method,it can be applied to robot grasping operations,auxiliary assembly monitoring,etc.
关键词
K-D树索引/BRIEF特征匹配/ICP点云配准/姿态估计Key words
K-D tree index/BRIEF feature matching/ICP point cloud registration/pose estimation引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据