近红外光谱对小麦产地来源的判别分析

Identification of Geographical Origins of Wheat with Discriminant Analysis by Near Infrared Spectroscopy

【目的】对不同产地来源小麦的近红外光谱进行判别分析,为小麦的产地鉴别提供一种新方法。【方法】应用近红外光谱分析仪检测2007/2008年度和2008/2009年度中国小麦主产区河北省、河南省、山东省和陕西省共240份小麦籽粒样品,对近红外光谱数据分别进行均值标准化、一阶求导和多元散射校正(MSC)处理后,利用偏最小二乘判别分析法(DPLS)分析预处理后的数据。【结果】2007/2008年度小麦籽粒样品总体正确判别率为87.5%,2008/2009年度样品总体正确判别率为91.7%;用2007/2008年度样品所建模型来预测2008/2009年度的样品,结果总体正确判别率为48.3%;两年样品混合后,总体正确判别率为82.5%。【结论】不同地域来源小麦的近红外光谱特征有显著差异,但其受品种和年际因素影响较大,判别模型的稳定性有待进一步提高。

【Objective】 A new method was developed for identification of the geographical origins of wheat with discriminant analysis by near infrared reflectance spectroscopy（NIRS）.【Method】A total of 240 wheat kernel samples collected in 2007-2008 and 2008-2009 from Hebei,Henan,Shandong and Shaanxi provinces of China were analyzed by NIRS.After normalization,first derivative and multiplicative scattering correction（MSC） pre-treatment of wheat kernel spectral data,discriminant partial least squares analysis（DPLS） was applied to classify the geographical origins of these samples.【Result】 There were 87.5% and 91.7% of wheat kernel samples collected in 2007-2008 and 2008-2009 were correctly classified respectively,the total correct classification of 48.3% was achieved by DPLS models developed using the samples collected in 2007-2008 harvest period to predict the geographical origin of the samples collected in 2008-2009 harvest period,and 82.5% were correctly classified by DPLS models developed using 2/3 of the total sample set to test the remaining 1/3 of the samples.【Conclusion】There are significant differences among near infrared spectra of wheat from different origins which are influenced mainly by the varieties and annual changes,and the stability of DPLS models developed has to be improved further.