Application of Near Infrared Spectroscopy and Pattern Recognition Model in Discrimination of Xihu Longjing Tea and Zhejiang Longjing Tea
近红外光谱和模式识别技术在西湖龙井与浙江龙井茶叶鉴别中的应用

To discriminate Xihu Longjing Tea and Zhejiang Longjing Tea,a recognition model was established by using a near infrared spectroscopy combined with chemometrics.First,the raw spectra were preprocessed with the Standard Normal Variant(SNV) function.Then,the Xihu Longjing Tea and Zhejiang Longjing Tea were predicted respectively with three models of Partial Least Square Regression- discriminant analysis(PLS-DA),Least Square Support Vector Machine(LSSVM) and Radial Basis Function Neural Network(RBFNN).The parameters of the LSSVM were optimized via grid search and leaveone -out cross-validation technologies.After optimization,the penalty coefficient(γ) and kernel function (δ~2) were 229.1 and 124.9 respectively.The best number of the neuron in hidden layer of RBFNN was 27. By comparison,LSSVM had the best prediction performance.Its RMSECV and R~2 in the calibration set were 0 and 1 respectively while its RMSEP and R~2 in the validation set were also 0 and 1 respectively. As a result,the Xihu Longjing Tea and Zhejiang Longjing Tea were classified correctly.