基于差异模型的印刷标签缺陷检测算法  

Printing Label Defect Detection Algorithm Based on Difference Model

作　　者：戴志威 林淼 陈超 胡钢 

机构地区：[1]深圳市研祥金码科技有限公司,广东深圳518057 [2]深圳市前海研祥亚太电子装备技术有限公司,广东深圳518057

出　　处：《工业控制计算机》2023年第6期63-64,共2页Industrial Control Computer

摘　　要：针对印刷标签的质量检测,提出一种基于差异模型的印刷标签缺陷检测算法,通过采集单张合格标签图像,分别进行高斯平滑、灰度腐蚀和灰度膨胀操作,获取多元模板图像,并用来训练差异模型。为了提高算法的鲁棒性,加入绝对阈值AbsThreshold与相对阈值Var Threshold参数,计算得到亮阈值图像和暗阈值图像。基于最小均方误差原理,求解出仿射变换参数向量,对待测标签图像进行配准对齐。最后通过差异模型,检测是否存在缺陷。该算法操作简单,对各种类型的印刷缺陷都能准确地检测出来。In order to detect the quality of printed labels,this paper proposes a printing label defect detection algorithm based on difference model.By collecting a single qualified label image,Gaussian smoothing,gray level corrosion and gray level expansion operations are carried out to obtain multivariate template images,and then used to train the difference model.In order to improve the robustness of the algorithm,the absolute AbsThreshold and relative VarThreshold parameters are added to obtain the bright-width value image and dark net value image.Based on the principle of least mean square error,the affine transformation parameter vector is solved,and the label image is registered and aligned.Finally,the difference model is used to detect whether there are defects.

关 键 词：差异模型 多元模板图像 仿射变换 图像配准 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TS896[自动化与计算机技术—计算机科学与技术]