Rhubarb Identification

Rhubarb is an important traditional medicinal plant. In Chinese pharmacopoeia, however, only three species are assigned as official rhubarbs: R. palmatem, R. tanguticum, and R. officeinale. Root and rootstalk of unofficial rhubarb were often mixed into official rhubarb in commercial products. But the medicinal function of unofficial rhubarb is not as good as the official rhubarb. In order to ensure the safety and curative effect of rhubarb in clinic practice, the discrimination of the official rhubarb is usually based on the shape, exterior configuration, microscopic features, and chemical components. But these methods can not be used to identify the powdered official and unofficial rhubarb samples.In this paper, back-propagation neural network (BP-ANN) and Takagi-Sugeno fuzzy system were applied to identify official rhubarb and unofficial rhubarb samples based on NIR spectra and mid-IR spectra. Neural networks are suitable for processing large amount of complicated data. The neural networks have some advantages over many other methods. Fuzzy logic is a mathematical method for solving inexact problems accurately. The T-S fuzzy system used in this paper is such a kind of fuzzy network. NIR spectra and mid-IR spectra were collected by using NIR spectrometers and mid-IR spectrometers (FOSS NIRSystem and Perkin Elmer, respectively).Results showed that these methods can be used to discriminate Chinese herbal medicine. The correctness of BP-ANN is 96% and the correctness of T-S is 100% by using NIR spectra of rhubarb. The correctness of BP-ANN is 98% and the correctness of T-S is 100% by using mid-IR spectra of rhubarb. In summary, the results obtained by Takagi-Sugeno fuzzy system are better than those by BP-ANN.