[Fast discrimination of varieties of bayberry juice based on spectroscopy technology]

Visible and near-infrared reflectance spectroscopy (NIRS) was applied in the discrimination of bayberry juice varieties. Characteristics of the pattern were analyzed by partial least square. Through full cross validation, nine principal components presenting important information of spectra were confirmed as the best number of principal components. Then, these nine principal components were taken as the input of BP neural network. Through the training and prediction, three different varieties of bayberry juice were classified according to the outputs of BP neural network. Besides, the work on building mathematic model and optimizing the algorithm was completed. In the process of BP neural network modeling, 60 samples were gained from the local market and each species has 20 samples. Fifty one samples were used as the training set and the reminder samples (total 9 samples) formed the prediction set. With a proper network training parameter, a 100% accuracy was obtained by BP neural network. Thus, it is concluded that PLS analysis combined with BP neural network is an available alternative for pattern recognition based on the spectroscopy technology.