基于近红外光谱的胡椒产地鉴别方法研究-《中国调味品》

文章信息/Info

Title:: Study on Identification Method of Pepper Origin Based on Near Infrared Spectroscopy

摘要:: 本研究旨在探索一种基于近红外光谱技术对胡椒产地进行分类的方法。收集海南、云南、广西、越南、马来西亚5个产地胡椒共计300份样品，采集近红外光谱。采用小波去噪等方法对光谱进行预处理，通过支持向量机(Support Vector Machine, SVM)、径向基神经网络(Radical Basic Function，RBF)和线性判别分析(Linear Discriminant Analysis，LDA)建立产地定性鉴别模型。研究表明，SVM和RBF神经网络模型鉴别准确率较好。db5小波预处理后的数据仅选择7个主成分正确率达到100%。结果表明基于近红外光谱的胡椒产地鉴别方法是可行的，预处理可以有效地提高近红外光谱胡椒产地鉴别模型的准确率。

Abstract:: This study aims to explore a method of classifying pepper producing area based on near infrared spectra. 300 pepper samples from Hainan, Yunnan, Guangxi, Vietnam and Malaysia were collected and the near infrared spectra of the samples were collected by the NIR analyzer. Using wavelet denoising and other methods of spectra preprocessing. Then support vector machine (SVM), radical basic function (RBF) neural networks and linear discriminant analysis (LDA) were applied to set up the identification models of pepper origins. The accuracies of the SVM model and RBF neural networks model were great. Data after pretreatment with db5 wavelet selected only 7 principal components, which classification accuracy is 100%.The results showed that it was feasible to identify pepper origins based on near infrared spectroscopy. And pretreatments could improve effectively the accuracies of identification models of the pepper origins.