基于近红外光谱技术的信阳毛尖产地判别  

Geographical Origin Discriminant of Xinyang Maojian Tea by Near Infrared Spectroscopy

作　　者：王子浩 刘洋[2] 李明玺 郭桂义 万春鹏[2] Wang Zihao;Liu Yang;Li Mingxi;Guo Guiyi;Wan Chunpeng(Henan Key Laboratory of Tea Plant Comprehensive Utilization in South Henan,Xinyang Agriculture and Forestry University,Xinyang,464000;Research Center of Tea and Tea Culture,College of Agronomy,Jiangxi Agricultural University,Nanchang,330045)

机构地区：[1]信阳农林学院,河南省豫南茶树资源综合开发重点实验室,信阳464000 [2]江西农业大学农学院,茶学与茶文化研究中心,南昌330045

出　　处：《分子植物育种》2019年第21期7161-7166,共6页Molecular Plant Breeding

基　　金：河南省豫南茶树资源综合开发重点实验室开放基金资助项目(HNKLTOF2017007);江西省教育厅科学技术研究项目(GJJ170291);江西农业大学教学改革研究课题(2017B2ZZ02);信阳农林学院2017年度青年教师基金项目(201701001)共同资助

摘　　要：信阳毛尖是极具地方特色的地理标志性产品,具有巨大的品牌价值。本研究基于近红外光谱技术(near infrared spectroscopy, NIRS)对信阳毛尖产地进行判别研究。采集4个不同产区信阳毛尖800~2 500 nm处的近红外光谱信息,并利用方差分析进行波长选择。选择了两组波长,同时对这两组波长进行数据拼接,分别进行产地的主成分分析(PCA);在拼接后近红外光谱的PCA图上,不同产地样品聚类最好。采用遗传算法(GA)在拼接的近红外光谱中选取特征波长。利用偏最小二乘法(PLS)分别建立以上选取的4组波长的关于信阳毛尖产地判别的分析模型,并采用留一法交叉验证其判别结果,结果表明基于遗传算法所获特征波长建立的产地判别模型是预测信阳毛尖产地能力最好的一种模型,其模型的预测值和计算的理论值之间呈现很好的线性关系,其相关系数为0.943 1,预测产地的准确率达到97.47%,具有较高的预测能力。本研究建立的NIR光谱结合PLS分析方法能够实现信阳毛尖产地的快速判别。Xinyang Maojian tea is a geographical product with local features, which has great brand value.Geographical origin discriminant of Xinyang Maojian tea by near infrared spectroscopy was studied. In this study,near-infrared spectral data from 800~2 500 nm of four different geographical origin Maojian tea were collected,then two sets of feature wavelengths were chosen andmerged by ANOVA. PCA was respectively used to distinguish the origin of Xinyang Maojian tea based on above three sets of wavelengths, and samples grouped best on PCA plot of merging spectral sequence. The characteristic wavelengths were selected by genetic algorithm(GA) using merging sequence. Partial least squares(PLS) were devoted to build the model for distinguishing quantificationally the quality grades of Maojian tea. The best result is obtained using characteristic wavelengths selected by GA, the theoretical values and predicted values showed good correlation, while the correlation coefficient is 0.943 1,prediction accuracy is 97.47%, which showed the accurate and reliable of the models. NIRS combined with PLS analysis were can be used for fast nondestructive testing the quality grades of Maojian tea.

关 键 词：近红外光谱 信阳毛尖 化学计量学 产地 

分 类 号：S571.1[农业科学—茶叶生产加工]