Fusing Spectral and Image Information for Characterization of Black Tea Grade Based on Hyperspectral Technology

Tea grade evaluation plays an important role in detecting the quality of tea. In this work, we proposed a method for black tea grade discrimination, which is based on the hyperspectral (HSI) technique, chemometric algorithms and data fusion strategies. Standard Normal Variable (SNV) and Multiple Scattering Correction (MSC) were used for the pretreatment of HSI-NIR data. Moreover, the spectral features of catechins, tea polyphenols and soluble sugars were extracted separately by the Uninformative Variable Elimination (UVE). To improve the accuracy of the model, the spectral data were secondarily downscaled by the Successive Projection Algorithm (SPA) and Kernel Principal Component Analysis (KPCA), respectively. Image features were extracted by Principal Component Analysis (PCA) and Gray Level Co-generation Matrix (GLCM). Simultaneously, the image features had been fused with the spectral features in order to realize data fusion. Partial Least Squares Discriminant Analysis (PLS-DA), Support Vector Machines (SVM) and Probabilistic Neural Networks (PNN) were utilized to build the discriminant models. The results show that hyperspectral techniques combined with data fusion strategies could improve the model accuracy. The UVE-SPA-GLCM-SVM was chosen as the optimal model with 98.33% accuracy of the validation set. The method provides a theoretical basis for online black tea grade differentiation.