微小针脚元件6D位姿估计  

6D Pose Estimation of Small Pin Component

作　　者：邝泳聪[1,2] 陈远波 罗益炜 刘东辉 KUANG Yongcong;CHEN Yuanbo;LUO Yiwei;LIU Donghui(School of Mechanical and Automotive Engineering,South China University of Technology,Guangdong Guangzhou 510640,China;Guangdong Key Laboratory of Precision Equipment and Manufacturing Technology,South China Univer-sity of Technology,Guangdong Guangzhou 510640,China)

机构地区：[1]华南理工大学机械与汽车工程学院,广东广州510640 [2]华南理工大学广东省精密装备与制造技术重点实验室,广东广州510640

出　　处：《机械设计与制造》2025年第4期113-120,共8页Machinery Design & Manufacture

基　　金：国家重点研发计划项目(2019YFB1310202)。

摘　　要：针对微小异形元件在被吸起时位姿容易发生偏转的问题,提出一种基于双目立体视觉和加权最小二乘的元件6D位姿估计算法。首先,提取图像上元件插装段针脚的中心点和末端中点并将其作为特征点,基于双目原理计算得到特征点的三维信息;其次,建立拟合元件6D位姿的加权最小二乘法模型,进而根据特征点的信息求得元件的6D位姿。最后,为验证算法的有效性,进行仿真实验和真实实验,结果表明:不同位姿下的针脚特征点在通过求得的6D位姿结果变换后,坐标标准差小于20μm,单次求解位姿耗时为540ms左右,算法精度及实时性均能满足元件的插装或分拣要求。Small odd-shaped component is prone to positional deflection when picked up.To solve this problem,a 6D pose estima⁃tion method which is based on binocular stereo vision and weighted least squares is proposed.Firstly,the center points and end midpoints of the component pins insertion part in images are extracted and used as feature points,and the 3D information of fea⁃ture points is calculated based on binocular principle.Secondly,a weighted least squares model for fitting the 6D pose of the compo⁃nent is established,and then the 6D pose of the component is obtained.Finally,to verify the proposed method,simulation and real experiment are conducted,and the results show that after using the obtained 6D pose result to transform,the standard deviation of the feature points'coordinates in different poses is less than 20μm.The time costs about 540ms.The accuracy and efficiency of the proposed method can meet the requirements for inserting or sorting small pin component.

关 键 词：机器视觉 双目视觉 最小二乘法 位姿估计 电子元件 

分 类 号：TH16[机械工程—机械制造及自动化] TP391.4[自动化与计算机技术—计算机应用技术]