Identification of Salvia Miltiorrhiza Radix et Rhizoma from Different Origins by Hyperspectrum with Artificial Neural Network

A method for identification of Salvia Miltiorrhiza Radix et Rhizoma from different origins was propsed by hyperspectrum with artificial neural network (ANN). The hyperspectra of Salvia Miltiorrhiza Radix et Rhizoma samples from 9 different origins were collected. The classification model for identification of origins of these samples was developed by using ANN combined with 5 different spectral preprocessing methods (maximum and minimum normalization, mean centralization, standard normal transformation, Savitzky-Golay smooth derivative and multiple scattering correction). As shown by the results of test set, when the number of hidden layer nodes was set to 17, the best model was developed by the mean centered method combined with ANN, with a classification accuracy of 98.77%. The sensitivity, precision and specificity of discrimination results from 7 Salvia Miltiorrhiza Radix et Rhizoma samples had reached 100.00%, the other 2 Salvia Miltiorrhiza Radix et Rhizoma samples were not less than 90.00%.