Nondestructive Variety Discrimination of Fragrant Mushrooms Based on Vis/NIR Spectral Analysis

The potential of artificial neural network (ANN) as a way for the nondestructive variety discrimination of fragrant mushrooms was evaluated. First, the visual and near infrared spectral data were analyzed by principal components analysis (PCA) for space clustering. The principal components (PCs) containing the main information of original spectra were picked out as the inputs of BP-ANN with three layers. The accumulative credibility of the first three PCs reaches 94.37%. The 3-D scores plot shows good space clustering of the samples. In the test, total 195 samples were examined, in which 150 samples were selected randomly for model-building and other 45 for model-prediction. With the prediction rate over 91%, the results indicate that the new hybrid model combing PCA with BP-ANN is reliable and practicable so that it could serve as an approach for machine recognition of various fragrant mushrooms.