Research on the grade detection of Lu'an Guapian tea based on MSC-SVM near-infrared spectroscopy

GUO Yidan

Abstract：In view of the cumbersome pre-processing of the existing near-infrared spectroscopy detection of crushed tea samples,this study utilizes the NIRS technology for rapid and non-destructive detection of the grade of Lu'an Guapian.In this paper,unprocessed dry tea samples were placed in ziplock bags,and the near-infrared spectroscopy of bagged tea samples were collected.Then,pretreated with multiplicative scattering correction(MSC),standard normal variate,first derivative(1st),Savitzky-Golay(SG)and the unprocessed spectral data were used,respectively.Support vector machine(SVM)algorithms was used to discriminate the quality of Lu'an Guapian.Experiments show that compared with the full-band model,the prediction accuracy of the 11,000～4,400 cm-1 band model is generally improved.Among them,the SVM prediction model established after MSC pretreatment had the best performance.The accuracy of the training set and the verification set are 93.33％ and 82.5％,respectively.Unknown samples are used to test the recognition effect of the models,and the accuracy of the test set of the model is 95.83％.The results showed that the MSC-SVM NIRS classification model of bagged teas based on 11,000～4,400 cm-1 band is the optimal model for grade discrimination of bagged tea,providing the best performance for tea quality detection.