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 oil-tea camellia seed oil by Vis /NIR spectra and UVE-GA method

作　　者：孙通[1] 胡田[1] 许文丽[1] 刘木华[1] 

机构地区：[1]江西农业大学生物光电技术及应用重点实验室,南昌330045

出　　处：《中国油脂》2013年第10期75-79,共5页China Oils and Fats

基　　金：国家自然科学基金项目(31271612);江西省教育厅科学研究基金(GJJ13254);江西农业大学科学研究基金(QN201105)

摘　　要：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％～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％。Pure oil-tea camellia seed oil and oil- tea camellia seed oils adulterated with 1% -10% of rapeseed oils were discriminated and classified by visible/near infrared （Vis/NIPt） spectra combined with uninformative variable elimination -genetic algorithm （UVE -GA） , and classification model was devel- oped by linear discriminant analysis （LDA）. The results indicated that UVE - GA was an efficient wave length variable selection method, and the classification model could be simplified and improved by UVE -GA. UVE -GA -LDA classification model was suitable for discriminating oil -tea camellia seed oil adulterated with more than 2% of rapeseed oil, and the correction rate of classification was 100% , while the correction rate of classification for oil - tea camellia seed oil adulterated with 1% of rapeseed oil was only 50% , which need to be improved.

关 键 词：可见 近红外光谱 UVE—GA 掺假鉴别 山茶油 

分 类 号：TS227[轻工技术与工程—粮食、油脂及植物蛋白工程] O657.33[轻工技术与工程—食品科学与工程]