[Fast discrimination of varieties of infant milk powder using near infrared spectra].

A new method for discrimination of varieties of infant milk powder by means of visible/near infrared spectroscopy (Vis/NIRS) (325-1075 nm) was developed. Partial least square (PLS) was used to analyze the characteristics of the pattern. PLS compressed thousands of spectral data into a small quantity of principal components and described the body of spectra. The first seven principal components were confirmed as the best number of principal components. Then, these seven principal components were applied as the input to a back propagation neural network with one hidden layer. The infant milk powder varieties data were applied as the output of BP neural network. One hundred eighty samples containing nine typical varieties of infant milk powder were selected randomly, and they were used as a training set of the BP neural network model, and the remainder samples (total 90 samples) formed the prediction set. With a proper network training parameter, the recognition accuracy of 100% was achieved. This model is reliable and practicable. So the present paper could offer a new approach to the fast discrimination of varieties of infant milk powder.