The Rapid Identification Of Radix Astragali From Different Regions By Multispectral Fusion-Sparse Representation Technology

Radix Astragali is one of a significant traditional Chinese medicine,which have pharmacological effects to various diseases.Due to the price and quality differences between Radix Astragali from different geographic origins and it is difficult to distinguish them from the form,thus establishing effective identification method of Radix Astragali from different geographic origins to improve the quality and management of Radix Astragali evaluation system is of great importance.Combining the instrument analysis method with the pattern recognition method is a research focus in the quality identification of Chinese herbal medicine such as Radix Astragali.Feature extraction method has been a research focus in the field of pattern recognition and an effective and robust feature extraction method is very important to improve the performance of classifiers.In this paper,kernel principal component analysis(KPCA)is used as a data feature extraction of different spectra data and it achieves high recogniton rate of Radix Astragali from different geographic origins.Under the effect of sparse representation classifier(SRC),experiments show that the data extraction of KPCA is effective in this paper.On the basis of 9-cross validation,the recognition rate increases from 59.99%to 94.51%,comparing with principal component analysis(PCA)combined with sparse representation classifier(SRC).Thus the core model of KPCA + SRC is established in this paper.In addition,in terms of single spectrum data,the data of ion mobility spectrum(IMS)is best for the classification of Radix Astragali from different geographic origins and two kinds of feature extraction maintain the consistency.Multispectral fusion technology is the current direction of instrument analysis,it combines the characteristics of different instruments to collect data for the comprehensive analysis of the sample.In this paper,the Raman spectroscopy(Raman),UV-Vis absorption spectroscopy(UV-vis)and IMS collect data of Radix Astragali from different regions.On the basis of two levels of hierarchy about the low level and data fusion technology,effective combination of different spectra data is given and under the model of KPCA + SRC,higher recognition rate data fusion method is presented.Experiments show that under the premise of two spectra data fusion,the middle level data fusion technology by UV-vis and IMS acheives the best effect and the highest recognition rate reaches 99.21%.The effect of the three spectra data fusion cannot beyond the middle level data fusion technology by UV-vis and IMS.In addition,the low level and middle level data fusion technology is analyzed and compared in this paper;Experiments show that low level of data fusion technology is more simple and higher recognition rate stability.The middle level data fusion technology can achieve higher recognition rate in some point of parameters than low level data fusion method,but the overall recognition rate is not stable,and it increases the algorithm complexity by choosing the combination of kernel principal components.Therefore,in terms of the data in this paper,the low level data fusion method is selected.The model of KPCA + SRC and multispectral fusion technology are combined to identify Radix Astragali from different geographic origins,which gets good identification results and it provides a feasible way for the astragalus quality and management evaluation system.