文献类型：期刊文献

中文题名：基于改进多类支持向量机的印刷缺陷检测

英文题名：Printing Defects Inspection Based on Improved Multi-Class Support Vector Machine

作者：胡方尚[1];郭慧[1]

机构：[1]华东理工大学机械与动力工程学院,上海200237

年份：2017

卷号：43

期号：1

起止页码：143

中文期刊名：华东理工大学学报（自然科学版）

外文期刊名：Journal of East China University of Science and Technology

收录：CSTPCD;;Scopus;北大核心:【北大核心2014】；CSCD:【CSCD_E2017_2018】；

语种：中文

中文关键词：缺陷检测;差分运算;支持向量机;动态阈值;印刷品

外文关键词：defect inspection;;subtraction operation;;SVM;;dynamic threshold;;printed matter

摘要：针对印刷品缺陷检测问题,为了对缺陷位置、形状、类型等信息进行有效的识别和分析,提出了一种基于改进多类支持向量机的印刷缺陷检测方法。首先根据人眼视觉特性,将配准后的印刷图像通过基于动态阈值的差分运算,快速地得到二值缺陷图像;然后采用由缺陷几何特征和形状特征构成的特征向量对缺陷信息进行分析和描述;最终通过改进的多类支持向量机实现印刷缺陷的准确识别。实验结果表明,相对于一对一型支持向量机(OVOSVM)和一对多型支持向量机(OVRSVM),在实际训练样本较少的情况下,该方法具有检测速度快、识别准确率高的特点,能够有效解决印刷品缺陷检测问题。
To recognize the defects of printed matter effectively,a method of printing defect inspection based on the improved multi-class support vector machine is proposed.According to the human visual characteristics,the binary defect image is rapidly obtained by the subtraction operation of registered image based on dynamic threshold.A feature vector consisting of defect geometric feature and shape feature is used to describe the defect of printing,and finally the accurate identification of printing defects is realized by the improved multi-class support vector machine.The experimental results show that in the case of less training samples the proposed method has faster detection speed and higher recognition accuracy than OVOSVM and OVRSVM,which can effectively solve the problem of printing defect inspection.