基于近红外光谱特征的国产绿茶产地溯源研究

Study on origin traceability of green tea based on near-infrared spectroscopy

中文关键词:  近红外光谱  绿茶  产区  溯源

英文关键词:near-infrared spectroscopy  green tea  producing regions  traceability

中文摘要:

      目的 研究我国多省出产绿茶近红外光谱特征, 探索以近红外光谱差异进行绿茶产地溯源研究的可行性。方法 运用傅里叶变换红外光谱仪漫反射法扫描我国多个产区的169份代表性绿茶样本, 谱图分别采用一阶导数(first derivative, 1st)、最大最小归一化(max-min normalization, Max-min)、标准正态变量变换(standard normal variable transformation, SNV)和多元散射校正(multiplicative scatter correction, MSC) 4种预处理方式处理, 采用连续投影算法(successive projections algorithm, SPA)筛选产地差异波段, 采用主成分分析(principal component analysis, PCA)对光谱降维处理, 建立茶叶产区判别模型。结果 采用Max-min、SNV、MSC处理的数据前3个主成分的累计方差贡献率均达到95%以上。分别建立基于主成分的线性判别分析(linear discriminant analysis, LDA)、支持向量机(support vector machine, SVM)、K-最近邻算法(K-nearest neighbor algorithm, KNN)判别模型。经交叉验证和测试集验证, 确定最佳处理路线为卷积平滑(Savitzky-Golay, S-G)+MSC+SPA+PCA, 结合KNN和SVM建立非线性模型, 其光谱测试集判别正确率均达到94%以上。结论 近红外光谱结合判别模型可以用于绿茶产地溯源, 本研究为绿茶溯源及模型参数选择提供了一定的参考。

英文摘要:

      Objective To study the near-infrared spectral characteristics of green tea produced in different provinces of China, and explore the feasibility on origin traceability of green tea by near-infrared spectroscopy. Methods The 169 representative green tea samples from multiple producing areas in China were scanned by Fourier transform infrared spectrometer diffuse reflection method. The spectra were preprocessed by 4 kinds of methods: First derivative (1st), Max-min normalization (Max-min), standard normal variable transformation (SNV) and multiplicative scatter correction (MSC). The successive projections algorithm (SPA) was used to screen the different bands of producing areas, and the principal component analysis (PCA) was used to reduce the dimension of the spectrum, and the discriminant model of tea producing areas was established. Results The cumulative variance contribution rate of the first 3 principal components of Max-min, SNV, MSC reached more than 95%. The linear discriminant analysis (LDA), support vector machine (SVM) and K-nearest neighbor algorithm (KNN) analysis based on principal component were established respectively. After cross-validation and test set verification, the optimal processing route was determined as Savitzky-Golay (S-G)+MSC+SPA+PCA, and the nonlinear model was established by combining KNN and SVM. The accuracy of the spectral test set was more than 94%. Conclusion Near-infrared spectroscopy combined with nonlinear discriminant model can be used for the origin traceability of green tea. This study provides a reference for the traceability of green tea and the selection of model parameters.