近红外光谱的主成分分析-马氏距离聚类判别用于卷烟的真伪鉴别  

Identification of Authentic and Fake Cigarettes Using Near Infrared Spectroscopy Combined with Principal Component Analysis-Mahalanobis Distance

作　　者：张灵帅[1] 王卫东[1] 谷运红[1] 邢军[2] 

机构地区：[1]郑州大学离子束生物工程省重点实验室,河南郑州450052 [2]国家烟草质量监督检验中心,河南郑州450001

出　　处：《光谱学与光谱分析》2011年第5期1254-1257,共4页Spectroscopy and Spectral Analysis

基　　金：国家自然科学基金项目(10505018)资助

摘　　要：为了快速准确的鉴别卷烟真伪,以A牌和假冒A牌卷烟为实验材料,采用近红外光谱法结合主成分分析-马氏距离判别分析方法建立了鉴别模型。首先对经过预处理的光谱数据进行主成分分析,分析结果表明,前4个主成分的累积贡献率已达98.46%,说明这4个变量能够代表原始光谱的主要信息。从120个样品中随机抽取100个用于建立4个主成分变量的定性判别模型,模型的相关系数达到了0.95,对20个未知样品的预测结果准确率为100%。说明近红外光谱结合模式识别方法进行卷烟真伪定性鉴别在技术上是可行的,可以作为卷烟真伪鉴别的一种辅助手段。In order to discriminate fake and genuine cigarettes correctly and rapidly,cigarettes of brand A and fake cigarettes of brand A were scanned by the NIR spectrometer,and an identifying model was developed by near infrared spectroscopy combined with principal component-Mahalanobis distance pattern recognition method.The pretreated spectra data of cigarette samples were analyzed through principal component analysis（PCA）,and the result of the analysis suggested that the accumulation of first 4 principal components was more than 97.46%.One hundred samples from total 120 cigarette samples were selected randomly.Then they were used to build qualitative discriminating model and the correlation coefficient was 0.95.Twenty unknown samples were validated by this model.The recognition rate is 100%.The model is reliable and practicable,and could be used as an assistant means for identifying fake and genuine cigarettes.

关 键 词：近红外光谱 卷烟 主成分分析-马氏距离 真伪 鉴别 

分 类 号：TS474[农业科学—烟草工业]