基于视觉的芯片位姿测量方法研究  

Research on Vision-Based Chip Pose Measurement Method

作　　者：冉迪 王宁 李松珍 王丹 韩志 王晓凤 RAN Di;WANG Ning;LI Songzhen;WANG Dan;HAN Zhi;WANG Xiaofeng(School of Mechanical Engineering,Shenyang Urban Construction University,Shenyang 110167,China;School of Mechanical Engineering,Shenyang University of Technology,Shenyang 110870,China;Engineering Training Center,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳城市建设学院机械工程学院,沈阳110167 [2]沈阳工业大学机械工程学院,沈阳110870 [3]沈阳工业大学工程实训中心,沈阳110870

出　　处：《组合机床与自动化加工技术》2024年第2期35-39,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金项目(52275455);辽宁省教育厅科学研究经费项目(LJKZ0160);辽宁省教育厅基本科研项目(JYTMS20231543)。

摘　　要：针对芯片贴装过程中的偏移检测问题,提出一种基于视觉的芯片位姿测量方法。该方法首先对获取的芯片图像进行预处理;在获取像素级边缘的基础上,根据像素点灰度值,利用Sigmoid函数进行分类,准确定位芯片亚像素边缘,并对边缘异常点进行校正;最后,由已知的芯片尺寸对图像中的连通域进行计数,得到图像中的芯片个数,并通过重心坐标将芯片亚边缘划分为4个区域,采用最小二乘法分别进行拟合,获得芯片的位姿信息。实验结果表明,该方法具有一定的鲁棒性,且定位精度在0.5 pixel以内,运行时间约为0.4 s,满足实际生产中的贴装要求,具有一定的使用价值。Aiming at the problem of chip offset detection during chip mounting,a vision-based chip pose measurement method is proposed.Firstly,the chip image is preprocessed.On the basis of pixel level edge,according to the gray value of pixel point,Sigmoid function is used to classify the chip sub-pixel edge accurately,and the edge outliers were corrected.Finally,the number of chips in the image is obtained by counting the connected components in the image according to the known chip size,the number of chips in the image is determined,and the chip sub-pixel edge is divided into four areas by the center of gravity coordinates,and the least square method is used to fit sub-pixel edge respectively,the pose information of the chip was obtained.The experimental results show that the method has certain robustness,and the positioning accuracy is less than 0.5 pixel,and the running time is about 0.4 s,which meets the requirements of mounting in actual production and has certain application value.

关 键 词：视觉检测 芯片定位 亚像素边缘 SIGMOID函数 异常点校正 

分 类 号：TH164[机械工程—机械制造及自动化] TG502[金属学及工艺—金属切削加工及机床]