不同预处理模式识别下的近红外绿豆产地溯源研究  

Traceability of NIR Mung Bean Origin under Different Pretreatment Modes

作　　者：胡亚光[1] 钱丽丽[1] 陈明明 李上 Hu Yaguang;Qian Lili;Chen Mingming;Li Shang(College of Food Science,Heilongjiang Bayi Agricultural University,Daqing163319)

机构地区：[1]黑龙江八一农垦大学食品学院,大庆163319

出　　处：《黑龙江八一农垦大学学报》2023年第1期59-64,共6页journal of heilongjiang bayi agricultural university

基　　金：战略性国际科技创新合作重点专项(2018YFE0206300,2018YFE0206300-0109)。

摘　　要：采用傅里叶变换近红外漫反射光谱仪测定来自吉林省白城市、黑龙江泰来县、黑龙江杜尔伯特蒙古自治县、山东省泗水县绿豆共120份样品的近红外光谱,分别采用一阶导数+9点平滑、标准正态变换(SNV)、多元散射矫正(MSC)、矢量归一化+MSC四种光谱预处理方法,建立偏最小二乘判别模型(PLS-DA),分析不同预处理方法对模型稳定性的影响,结果得出:原始光谱模型判别率为62.5%,一阶导数+9点平滑预处理模型判别率为65%,SNV预处理模型判别率为65%,MSC预处理模型判别率为82.5%,矢量归一化+MSC预处理模型判别率为90%。因此,采用矢量归一化+MSC预处理方法对绿豆产地判别的准确率最高。A Fourier transform near-infrared diffuse reflection spectrometer was used to determine the near-infrared spectra of 120samples from Baicheng city,Jilin province,Tailai county,Heilongjiang province,duerbert Mongolia Autonomous county,Heilongjiang province and Sishui county,Shandong province.The first derivative+9-point smoothing,standard normal transformation(SNV),multivariate scatter and Vector-normalized+MSC were used respectively.The results showed that the discrimination rate of original spectral model was 62.5%,the discrimination rate of first derivative+9-point smoothing pretreatment model was 65%,the discrimination rate of SNV pretreatment model was 65%,and the discrimination rate of MSC pretreatment model was 82.5%The discriminant rate of vector normalization+MSc pretreatment model was 90%.Therefore,vector normalization+MSC pretreatment method of mung bean origin discriminant had the highest accuracy.

关 键 词：绿豆 近红外光谱 预处理方法 产地溯源 模型稳定性 

分 类 号：S522[农业科学—作物学] TS201.6[轻工技术与工程—食品科学]