《基于UVE-GA变量优选的山茶油可见近红外光谱掺假鉴别 Adulteration discrimination of off-tea camellia seed oil by VisNIR spectra and UVE-GA method》.pdfVIP

2013年第38卷第10期 中国油脂 75 基于UVE—GA变量优选的山茶油 可见／近红外光谱掺假鉴别 孙通，胡 田，许文丽，刘木华 (江西农业大学生物光电技术及应用重点实验室，南昌330045) 摘要：利用可见／近红外光谱结合无信息变量消除一遗传算法(UVE—GA)变量选择方法对山茶油 和掺杂低比例菜籽油(1％一10％)的山茶油进行鉴别分类，并应用线性判别分析(LDA)方法建立 分类模型。结果表明：UVE—GA是一种有效的波长变量选择方法，能简化分类模型和提高分类模 型精度；UVE—GA—LDA分类模型适用于掺杂2％以上菜籽油的山茶油鉴别分类，其分类正确率 为100％；对掺杂1％菜籽油的山茶油鉴别分类正确率有待提高，其分类正确率仅为50％。 关键词：可见／近红外光谱；UVE—GA；掺假鉴别；山茶油 中图分类号：TS225．1；0657．3文献标志码：A Adulterationdiscriminationofoil—．teacamelliaseedoil by Vis／m andUVE—GAmethod spectra SUN Muhua Tian，XUWenli，LIU Tong，HU ofBiomaterials (Optics—ElectronicsApplication Lab，JiangxiAgriculturalUniversity， Nanchang330045，China) oil—teacamelliaseedoilandoil—teacamelliaseedoilsadulteratedwith1％一10％of Abstract：Pure oilswerediscriminatedandclassifiedvisible／near combinedwith rapeseed by infrared(Vis／NIR)spectra wasdevel- uninformativevariable classificationmodel elimination—geneticalgorithm(UVE—GA)，and lineardiscriminant resultsindicatedthatUVE—GAwasanefficientwave opedby analysis(LDA)．The variableselection theclassificationmodelcouldbe and length method，and simplifiedimprovedby classificationmodelwassuitablefor oil—teacamelliaseed UVE—GA．UVE—GA—LDA discriminating was oiladulteratedwithmoretha