基于标准互相关函数的图像分辨率自动标定方法  

A Method of Image Resolution Calibration for Automatic Microscopic Zooming Based on NCCO

作　　者：孙明磊[1] 宗光华[1] 董代[2] 石晶欣[1] 

机构地区：[1]北京航空航天大学机械工程及自动化学院,北京100083 [2]北京航天控制仪器研究所,北京100039

出　　处：《光学学报》2008年第6期1117-1123,共7页Acta Optica Sinica

基　　金：国家863计划(2004AA404260);中国博士后科学基金(20070420287)资助课题

摘　　要：理想规则图像特征经标准互相关函数匹配后,相似函数C*(x)可用确定的解析式Z(x)表达。但是对于含噪声图像,相似函数C(x)较之C*(x)发生了变化,但存在"零相似不变性"。将此原理应与于显微视觉中图像物面分辨率的在线标定,推导了矩形图像特征的一维相似函数解析式Z(x);求解C(x)=0作为Z(x)的近似,并给出了具体标定算法;通过仿真图像实验,给出标定算法的正确度在0.1～0.2pixel;最后,将标定算法应用于可连续变焦的微对准装配系统,实测算法的精密度可达为0.08pixel。实验结果表明,基于"零相似不变性"的标定算法是具有实用性的亚像素标定方法。Pattern matching tool of normalized cross correlation operator（NCCO）computation C^＊（x）for ideal images can be determined by analytical expression Z（x）.Compared with C^＊（x）,C（x）of NCCO matching between images containing noise shows a little difference.But there is a property of zero similarity invariance（ZSI）.A method of image resolution calibration in microscopic vision was proposed.Firstly,similarity functions of rectangular feature in x direction Z（x）were derived and proved.Then C（x）=0 was solved as an approximation of solution of Z（x）.The detailed ZSI based calibrating algorithm was illuminated.The results of simulating image experiments show that the calibration accuracy of the algorithm can reach 0.1～0.2 pixel.Finally,the practical experiments were conducted on a hybrid microassembly workcell.Experimental results show that the calibration precision of the algorithm can reach 0.08 pixel.The ZSI based calibrating algorithm is a method with practical value and subpixel precision.

关 键 词：信息光学 显微视觉 图像 标定 标准互相关 

分 类 号：TP242.62[自动化与计算机技术—检测技术与自动化装置]