Classification Of Tea Based On Fourier Transform Infrared Spectroscopy

Tea contains tea polyphenols,caffeine and soluble solids,which are beneficial to the health of the body.For example,tea polyphenols can inhibit atherosclerosis,reduce the incidence of heart head blood-vessel.Caffeine can increase the secretion of gastric juice,promote digest,etc.,therefore,tea became one of the most popular health drink.There are many kinds of tea.It is difficult to distinguish them by appearance.In this paper,using Fourier Transform Infrared Spectroscopy(FTIR)combined with pattern recognition statistical analysis method for the classification and identification of tea.Firstly,using the unsupervised learning method of fuzzy set theory,the linear discriminant analysis is extended to the fuzzy algorithm,and a fuzzy Fisher linear discriminant analysis algorithm is developed.This algorithm can get a set of optimal unrelated discriminant vector.This set of optimal unrelated discriminant vector not only satisfies the Rayleigh quotient equation,also satisfies the projected non-correlation of the sample to the fuzzy uncorrelated discriminant vector.This paper provides a basis for the subsequent study of the Fourier Transform Infrared spectral classification model.Secondly,developed a fuzzy uncorrelated discriminant c-means clustering model and a fuzzy discriminant c-means clustering model.The two models are based on Fourier Transform Infrared Spectroscopy,principal component analysis,fuzzy fisher linear discriminant analysis and fuzzy c-means clustering.The classification of three kinds of tea varieties was analyzed by using this two models during the second-order derivative spectra in the range of 4000~400cm-1.Results show that the accuracy of fuzzy identification c-means model is 95.45%,which is higher than the accuracy of fuzzy uncorrelated identification c-means clustering model.Finally,in order to study a higher accuracy and faster convergence method of tea classification,the fuzzy membership of the fuzzy c-means algorithm is used as the learning rate.Making the fuzzy c-means clustering algorithm integrate into the kohonen clustering network algorithm to establishing a fuzzy differential kohonen cluster analysis model.The model can not only automatically control the learning rate distribution and update the neighborhood,but also can extract the information and compress the data dimension dynamically during the clustering process.The tea varieties classification accuracy is99.94%.The results show that the model used in FTIR can identify the three different varieties of tea rapidly and accurately.The research results show that the FTIR combined with pattern recognition analysismethod can identify the different varieties of tea quickly and accurately,for the classification of tea identify research provides a rapid and accurate method.