近红外光谱组合化学计量学用于香烟品牌在线识别  

The Combination of Near-infrared Spectroscopy and Chemometrics for Online Distinguishing Different Brands of Cigarettes

作　　者：谭超[1,2] 吴同[1] 许泽宏[1] 李惟一[1] 张开仕[1] 覃鑫 

机构地区：[1]宜宾学院化学与化工学院,四川宜宾644000 [2]宜宾学院计算物理四川省高校重点实验室,四川宜宾644000 [3]川渝中烟,成都610017

出　　处：《宜宾学院学报》2011年第12期65-67,73,共4页Journal of Yibin University

基　　金：四川省青年科技基金(09ZQ026-066);宜宾学院科研与教学创新团队资助项目(Cx201104);宜宾学院重点项目(2011Z22)

摘　　要：用组合近红外光谱和化学计量学方法——相似分类法(SIMCA)和支持向量机(SVM)——来识别不同品牌烟草产品.通过一个案例研究发现,当训练样本较少时,SVM分类器优于SIMCA分类器;而当训练样本相对充足时,两类分类器精度无显著差异;SVM分类器对训练集组成依赖较小.结果表明:支持向量机组合近红外光谱是一个烟草质量控制中有价值的工具,可改变长期以来烟草产品的质量评估依赖于专家的感官这一传统方法,为评估结果的准确性和公正性提供有力手段.The feasibility of combining near-infrared(NIR) spectroscopy and chemometrics,i.e.,soft independent modeling of class analogies(SIMCA) and support vector machine(SVM),for discriminating different brands of cigarettes was investigated.A case involving three different brands of cigarettes reveals that the accuracy of SVM classifiers is obviously higher than that of SIMCA when the training set is small;whereas,the training set is relatively large,both kinds of classifiers show comparable results.Moreover,the SVM classifier depends less on the composition of the training set.Those results suggest that the combination of NIR spectroscopy and SVM is a valuable alternative for quality control of cigarettes.It helps to change the fact that the discrimination of cigarette quality has long been dependent on the expert sensory evaluation and also provides a powerful tool for ensuring the accuracy and the fairness of the evaluation.

关 键 词：近红外光谱 香烟 品牌识别 支持向量机 

分 类 号：O65[理学—分析化学]