Fast Nondestructive Identification of Endothelium Corneum Gigeriae Galli Using Visible/Near-Infrared Spectroscopy

摘要

Vis/NIR, spcctroscopy, in combination with partial least square (PLS) analysis and a back-propagation neural network, is investigated to identify cndothcliutn corncum gigeriae galli (EC'GG). The spectra] features of ECGGs and their counterfeits are reasonably differentiated in vis/NfR region, whicli provides enough qualitative information to establish the relationship between the spectra and samples for identification. After pretrcatment of the spectral data, cross validation is implemented for extracting the frest number of principal components. Then the calibration and validation set arc performed well. The PLS and back propagation neural network (BPNN) model gives the BPNN to be 0.9941 and the root mean square residual (RMSR) to be 0.0775 for the calibration set, and the multiple correlation coefficient (MCC) to 0.9874 and the RMSE to 0.1134 for the validation set. Thus the PLS and BPNN model is reliable and practicable. Through testing, a recognition accuracy of 100% is achieved. The present study could offer a new approach for fast and nondestructive discrimination ofECGG and its counterfeit.%Vis/NIR spectroscopy,in combination with partial least square(PLS)analysis and a back-propagation neural network,is investigated to identify endothclium corneum gigeriae galli(ECGG).The spectral features of ECGGs and their counterfeits are reasonably differentiated in vis/NIR region,which provides enough qualitative information to establish the relationship between the spectra and samples for identification.After pretreatment of the spectral data,cross validation is implemented for extracting the best number of principal components.Then the calibration and validation set are performed well The PLS and back propagation neural network(BPNN)model gives the BPNN to be 0.9941 and the root mean square residual(RMSR)to be 0.0775 for the calibration set,and the multiple correlation coeffcient(MCC)to 0.9874 and the RMSE to 0.1134 for the validation set.Thus the PLS and BPNN model is reliable and practicable.Through testing,a recognition accuracy of 100％ is achieved.The present study could offer a new approach for fast and nondestructive discrimination of ECGG and its counterfeit.