关键词: 黄芪, 傅里叶变换红外光谱, 指纹图谱, 正交偏最小二乘-判别分析, 反向传播-人工神经网络

Abstract: In this study,collecting the fingerprints of 90 batches of Astragali Radix from 12 counties,investigating the spectral characteristics of Astragali Radix from different origins,and establishing model of OPLS-DA (orthogonal partial least squares-discriminant analysis) and BP-ANN (back-propagation neural network) to realize the rapid identification and prediction of origin.The results showed that the FTIR fingerprints and its average spectrums of Astragali Radix from different origins showed some differences;the established OPLS-DA model can be used to accurately classify Astragali Radix from different origins;the differences of FTIR fingerprints of Astragali Radix from different origins were mainly existing in the wave number ranges of 3400-2920,1650-1420,1370-1050,930-770 cm^-1,and 590-520 cm^-1;the prediction accuracy of testing set was higher than 83% based on BP-ANN model.Hence the established OPLS-DA and BP-ANN model based on FTIR fingerprints can be used to identify and predict the origin of Astragali Radix.

Key words: Astragali Radix, fourier transform infrared spectroscopy, fingerprint, orthogonal partial least squares-discriminant analysis, back-propagation neural network