傅里叶变换红外光谱技术鉴别冬虫夏草真伪的研究  

Authentication of Ophiocordyceps sinensis by Fourier Transform Infrared Spectroscopy

作　　者：张九凯[1] 张小磊[1,2] 曾文波[3] 邹小波[2] 虞泓[3] 陈颖[1] 

机构地区：[1]中国检验检疫科学研究院,北京100123 [2]江苏大学食品与生物工程学院 [3]云南大学中草药生物资源研究所

出　　处：《检验检疫学刊》2015年第3期1-7,共7页Journal of Inspection and Quarantine

基　　金：国家"863"计划项目(2011AA100807);中国检验检疫科学研究院基本科研业务费专项资金资助项目(2014JK021)

摘　　要：研究了傅里叶变换红外光谱技术结合模式识别对冬虫夏草进行真伪鉴别的可行性。收集不同产地的野生冬虫夏草和常见伪品,根据各样品在4 000-450 cm^(-1)范围内的红外光谱,确定冬虫夏草指纹区域所在的位置,讨论不同预处理方法及不同光谱区间对建模结果的影响;同时利用主成分分析将前15个主成分作为输入变量建立冬虫夏草真伪鉴别的模式识别模型。结果表明:冬虫夏草红外光谱指纹区为1 800-800 cm^(-1),最佳预处理方法为多元散射校正;指纹区光谱经多元散射校正预处理和主成分分析降维后,所建立的模式识别模型中支持向量机模型的训练集和交互验证集的识别率均能达到100%。The feasibility of authentication of Ophiocordyceps sinensis(DCXC) based on Fourier transform infrared spectroscopy(IR) and pattern recognition was studied.For currently different adulteration ways on the market,samples of genuine DCXC,its counterfeits(other species of Cordyceps,CC),cultured mycelia(Cordyceps flower,CCH) and mimics(Stachys sieboldii,CSC) were selected,and the infrared spectrum in the region of 4 000-450 cm^(-1) was collected afterwards.The IR fingerprint region of DCXC was determined,and the effects of different pretreatment methods and spectral ranges on modeling results were discussed.After dimensionality reduction by principal component analysis(PCA),the former 15 principal components were used as input variables to build authentication models using different validated pattern recognition methods.Results showed that fingerprint region of DCXC was 1800-800 cm^(-1),and the optimum pretreatment method was multiplicative scatter correction(MSC).Recognition rate of training set and cross-validation set of support vector machine model based on the fingerprint region can both reach100%after MSC pretreatment and PCA.

关 键 词：红外光谱 模式识别 多元散射校正 主成分分析 支持向量机 

分 类 号：O657.33[理学—分析化学] R282.5[理学—化学]