基于Hough变换的高精度机器视觉对准系统的研究  

High-precision alignment system of machine vision based on Hough transform

作　　者：魏丽[1] 郭玉静[1] 陆向宁 

机构地区：[1]唐山学院机械工程省级实验教学示范中心,河北唐山063000 [2]江苏师范大学机电工程学院,江苏徐州221116 [3]华中科技大学数字制造装备与技术国家重点实验室,湖北武汉430074

出　　处：《应用光学》2017年第6期884-889,共6页Journal of Applied Optics

基　　金：国家自然科学基金(51305179);数字制造与装备技术国家重点实验室开放基金(DMETKF2016005)

摘　　要：随着集成电路特征尺寸的不断减小,I/O引脚数日益增加。对于高密度封装,尤其是2.5D/3D封装中,电子元件和基底快速而准确的对准十分关键。利用高性能CCD搭建了一套精密的光学视觉对准系统,并采用Hough变换算法对芯片和基底进行对准试验。采集的图像经过去噪预处理后采用形态学边缘检测方法提取对齐标记,并通过Hough变换得到4组平行直线,然后计算基底的位移和旋转角度并完成对准。使用Matlab编程配准,程序运行时间约为4.2s,旋转参数的误差小于1.2°,x和y轴的平移误差均小于1pixel。试验结果表明基于Hough变换算法的光学视觉对准系统可以快速而精确地实现芯片和基底对准,满足IC封装需求。As the integrated circuit（IC） characteristic size continues to decrease, the number of I/ O pins is increasing. For high-density packaging, especially in 2.5D/3D packages, fast and accurate alignment of electronic components and substrates is critical. A precise optical vision alignment system was set up by using the high precision charge coupled device （CCD） , and the Hough transform algorithm was utilized to calibrate the chip and the substrate. The image captured was preprocessed to remove the noise, and the alignment marks were extracted by using morphological edge detection method, and 4 sets of parallel straight lines were obtained by Hough transform. Finally, the displacement and the rotation angle of the substrate were calculated and the alignment was completed. The running time of the Matlab program is about 4.2 s, the rotation parameter error is less than 1.2 °,x and y axis translation errors are less than 1 pixel. The experimental results demonstrate that the optical visual alignment system based on the Hough transform algorithm can realize chip and substrate alignment quickly and accurately to meet the IC packaging requirements.

关 键 词：对准系统 霍夫变换 机器视觉 倒装芯片 

分 类 号：TN206[电子电信—物理电子学]