基于支持向量机的条烟包装外观缺陷检测  

Appearance Defect Detection of Cigarette Packaging Based on Support Vector Machine

作　　者：孙娜 管一弘[1] 崔云月 罗亚桃 黄岗 SUN Na;GUAN Yi-hong;CUI Yun-yue;LUO Ya-tao;HUANG Gang(School of Science,Kunming University of Science and Technology,Yunnan 650500,China;Kunming Julin Technology Company Limited,Yunnan 650000,China)

机构地区：[1]昆明理工大学理学院,云南昆明650500 [2]昆明聚林科技有限公司,云南昆明650000

出　　处：《软件》2020年第1期205-210,共6页Software

摘　　要：针对卷烟生产过程中条烟包装外观缺陷问题,提出一种基于支持向量机(SVM)的条烟包装缺陷图像检测方法。该方法首先采用模板匹配法定位条烟检测区域;然后利用Haar小波变换进行频域分解,并通过灰度共生矩阵算法对频域图提取纹理特征;最后结合纹理特征建立条烟支持向量机分类模型,对待测样本进行分类识别。结果表明:基于SVM分类模型的识别率为96.1%,该方法通用性强,实时性好,满足条烟异常情况检测要求。与BP神经网络测试性能相比,分类性能优于BP神经网络。Aiming at the appearance defects of cigarette packaging in cigarette production process, an image detection method of cigarette packaging defects based on support vector machine(SVM) is proposed. Firstly, the template matching method is used to locate the cigarette detection area. Then, Haar wavelet transform is used to decompose the frequency domain, and texture feature is extracted from frequency domain image by gray level co-occurrence matrix algorithm. Finally, a support vector machine classification model is established based on texture feature, and the samples to be tested are classified and recognized. The results show that the recognition rate based on SVM classification model is 96.1%. The method is universal and real-time, and meets the detection requirements of abnormal situation of cigarettes. Compared with BP neural network, its classification performance is better than BP neural network.

关 键 词：图像处理 缺陷检测 支持向量机 条烟 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]