Research Abont The Methods Of Detecting Contents Of Main Components And Discriminating Varieties Based On NITS

The content of protein and fat is the main index for measuring the quality of milk and dairy. Currently, the content of protein and fat in the milk and dairy are detected by the conventional chemical methods at home and abroad, But the methods exist some shortcomings, such as the high cost of equipment, slower detection, destruction of the sample and complex operation, which can’t meet the high-speed, continuous and automatic development needs of dairy industry. Then, near infrared spectroscopy gets more and more attention for the advantages of being quick and nondestructive.In this study, four kinds of liquid dairy products in Ningxia local were taken as the experimental subjects, the near infrared transmission spectroscopy (NITS) technology was used, wavelet transform (WT) and artificial neural networks (ANN) were combined, the prediction models of content of protein and fat as well as the species identification models were established respectively. Then, the results of the models were evaluated and analyzed. The purpose was to find a new method for detecting the quality of dairy quickly and nondestructively as well as to provide an important basis for online testing of the dairy products. The main research achievements of the paper are as follows:(1)The pure milk, sour milk, wheat milk and wolfberry milk four kinds of milk of a total of298samples from different production batches were selected for experiment, and the chemical standard values of the content of protein and fat in the samples were measured.(2)The prediction models of the content of protein and fat in the samples were established by the PLS, BP-ANN, RBF-ANN respectively, and the best result was to predict the wheat milk based on the RBF-ANN. When the protein was modeled, the correlation coefficient R and the prediction mean square error RMSEP of pure milk, sour milk, wheat milk and wolfberry milk were0.99990.03010.99720.060.99960.04120.99970.0331respectively,when the fat was modeled, the R and RMSEP of the four kinds of milk were0.99970.09680.9980.06880.99950.07390.99960.065respectively. By the feedback of the results of modeling, we get that the best spectral preprocessing methods of protein and fat were SD+S-G smoothing and FD+MSC+S-G smoothing respectively.(3)240samples from the four kinds of dairy were selected for experiment.60samples were selected from each kind of dairy. Then the variety identification models of the dairy were established by the BP-ANN, RBF-ANN and SVM-ANN respectively. The identification rate to the three models was100%with MSC+SG+WT method for spectral preprocessing. Overall, the best identification result was based on the SVM-ANN model.The results showed that the NITS combined with chemometric methods could achieve rapid and nondestructive testing of dairy quality as well as variety identification. It could provide a theoretical basis for the online testing platform of domestic dairy component analysis.