基于序列局部图像的高精度测量  

A high-precision dimension measurement method based on sequential partial images

作　　者：何博侠[1] 张志胜[1] 戴敏[1] 史金飞[1] 

机构地区：[1]东南大学机械工程学院,江苏南京211189

出　　处：《光学精密工程》2008年第2期367-373,共7页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.70272046)

摘　　要：研究了基于序列局部图像的视觉测量方法。分析了机械零件图像边缘的过渡分布特征,提出用边缘像素补偿法来消除实际边缘不能精确定位对测量精度的影响。以直线边缘距离测量为原型,提出了基于序列局部图像尺寸特征的测量方法:对零件进行微小区域成像,生成在空间上连续的序列局部图像;应用相关系数法和双线性插值法获得相邻序列图像的亚像素级尺寸特征线,从而得到各局部图像的尺寸特征;对这些尺寸进行求和与补偿,得到零件的总体尺寸。实验结果表明,对常规尺寸零件的单幅图像运用边缘像素补偿法,相对测量误差在0.008%以内;对大尺寸零件应用序列图像测量法,相对测量误差在0.01%以内,并具有误差积累小的优点。本文算法可基本满足板类零件的精密自动化测量要求。A vision measurement method based on dimension characteristics of sequential partial images is proposed. The transition distribution modality of part edge image and its effect on measurement accuracy are analyzed. To get rid of the effect, an edge pixel compensation method is proposed to availably improve measurement accuracy. By taking the dimension measurement of straight edge parts as prototype, the vision measurement method based on dimension characteristics of sequential partial images is introduced. This method consists of three steps：the small areas of the part are imaged and then the sequential partial images are overlapped at adjacent margins one by one. The cross-correlation matching technique and the bilinear interpolation algorithm are applied to obtain the dimension characteristic lines of the adjoining sequential images at sub-pixel level and the corresponding dimension characteristics of each partial image, and these sizes are summed and compensated to get the part dimension. The experimental results show that for the edge pixel compensation method to the single image of general scale part, the relative error is less than 0. 008 %,and for the sequential image measurement method to the large scalar part, the relative error is less than 0.01%, which indicates that the proposed method has the advantage of small error accumulation and can meet the needs of the precise auto measurcment of sheet metal parts.

关 键 词：视觉测量 图像处理 序列图像 边缘补偿 特征匹配 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TG806[自动化与计算机技术—计算机科学与技术]