Nondestructive Discrimination of Peach Varieties Using Near Infrared Spectroscopy

A new method for the discrimination of peach varieties by Vis/near infrared reflectance spectroscopy (NIRS) (325-1075nm) was developed. A relationship has been established between the reflectance spectra and peach varieties. The set was consisted of a total of 90 samples of peaches. First, the data was analyzed by principal component analysis (PCA). PCA compressed thousands of spectral data into a small quantity of principal components, which described the main information of the spectra. Then the first 8 principal components from PCA were used as inputs of a back propagation neural network with one hidden layer. 75 samples were selected randomly from three varieties and used as the training set to build BP-ANN model. This model was used to predict the varieties of 15 unknown samples. The residual error of regression was 2.431times10^-3, and the accuracy was 100%. The result of the PCA-BPNN method is much better than that of the PCA method. It is concluded that Vis/NIRS is a good method for the discrimination of peach varieties based on PCA-BPNN