基于双显微视觉的光刻版位姿定位算法  

Positioning Algorithm of Lithographic Mask Posture Based on Binocular Micro-vision

作　　者：吕红阳 陈立国 仲丁元 薛立伟 LüHong-yang;CHEN Li-guo;ZHONG Ding-yuan;XUE Li-wei(Robotics & Microsystem Center, School of Mechanical and Electrical Engineering, Soochow University, Suzhou 215021, China)

机构地区：[1]苏州大学机电工程学院机器人与微系统研究中心,苏州215021

出　　处：《科学技术与工程》2020年第32期13295-13301,共7页Science Technology and Engineering

基　　金：江苏省高校自然科学研究项目(17KJA460008)。

摘　　要：光刻版工件移动位姿的精确定位及调整是完成其视觉对位的关键。针对光刻版工件跨尺度级高精度对位的难点,搭建了双显微视觉对位系统,通过双显微视觉的局部位姿检测,解决了工件尺寸与定位精度之间的矛盾。具体方法为:首先采用双显微视觉获取工件两端局部图像;接着提出改进Canny算法并基于多项式插值的边缘细分完成亚像素级边缘轮廓提取;然后基于RANSAC算法拟合边缘轮廓,获得左右相机图像中“Mark”标志中心点位置坐标及偏转角度;最后通过推导局部位姿间数学关系完成光刻版工件的精确位姿定位。实验结果表明:所设计的视觉算法对于0.5 mm平行线的距离检测精度达1.93μm,角度提取精度达0.018°;对于光刻版的移动位姿的定位精度达0.64μm,能满足视觉对位过程中高精度的定位需求。The precise positioning and adjustment posture of the moving pose of the lithographic masks is the key to successfully completing visual alignment.According to the difficulty of cross-scale high-precision alignment of lithographic mask,a binocular micro-vision alignment system was built.The contradiction between the masks size and positioning accuracy was resolved by detecting partial pose.The specific method was shown as follows.Firstly,the partial images at the ends of the mask were obtained by double micro-vision.Then,the sub-pixel edge profile extraction was completed based an improved Canny algorithm and polynomial interpolation subdivision.Based on the RANSAC algorithm to fit the edge contour to obtain the left and right cameras,the edge profile was fitted based on the RANSAC algorithm to obtain the center point position and deflection angle of the mask"Mark"in the left and right camera images.Finally,the precise posture of the lithographic mask could be achieved by deducing the mathematical relationship between the partial pose.The experimental results show that for 0.5 mm parallel lines the distance accuracy of designed vision algorithm is 1.93μm and the angle accuracy is 0.018.The positioning accuracy of the lithographic mask's moving posture can reach 0.64μm,which can meet the high-precision positioning requirements in the process of visual alignment.

关 键 词：视觉对位 亚像素边缘检测 RANSAC算法 位姿检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]