Discrimination of Varieties of Dry Red Wines Based on Independent Component Analysis and BP Neural Network

In order to achieve the rapid discrimination of the varieties of red wines, the authors selected 5 kinds of dry red wine for study with Vis/NIR spectroscopy. Firstly, Characteristics of the pattern were analyzed by independent component analysis (ICA). Through comparing the results of modeling performance by different number of independent components, 20 principal components presenting important information of spectra were confirmed as the best number of principal components. The 20 independent components (ICs)extracted by ICA were employed as the inputs of the BP neural networks, and then a three layers of BP neural network was built, category analysis was performed, and the work of building mathematics model and optimizing the algorithm was completed. Five samples from each variety and a total of 25 samples were selected randomly as the prediction sets. The remaining 150 samples were used as the training sets to build the training model, which was validated by the samples of the prediction sets. The recognition rate was 100%. In addition, based on the independent component analysis, the authors selected two characteristic wave bands in reference to vector loading map of mixed matrix. So the pattern recognition methods developed in this paper not only played a good role in the classification and discrimination, but also had the capability to extract the finger feature of red wine, and offered a new way for detecting and developing red wines.