Discriminating varieties of tea plant based on Vis/NIR spectroscopy and chemometrics

A novel method for discriminating the varieties of tea plant based on Vis/NIR spectroscopy was developed. Firstly, the Vis/NIR reflectance spectra of leaves were measured by spectroradiometer (handheld, FieldSpec) nondestructively at 325-1075 wavelength range in tea garden. Secondly, the spectral data were pretreated to eliminate the system noises and disturbances. Thirdly, the chemometrics was used to extract the diagnostic information and build the discrimination model. The chemometrics was integrated with the wavelet transform (WT), principal component analysis and artificial neural networks (ANN). WT was used to reduce data dimension and mine information from spectra data. The diagnostic information from WT could be re-expressed and visualized in principal components (PCs) space, which could lead to discover the structures correlative with the different varieties. The first eight PCs, which accounted for 99.29% of the original variables, were used as the input of the ANN model. The ANN model yielded acceptable classification accuracy with the proper spectral pretreatment and optimum WT parameter. The discrimination accuracy was 86.7% for the varieties of samples in the prediction set. It can be concluded that the varieties of tea plant can be discriminated by using Vis/NIR reflectance spectroscopy.