Detection Method Of Chlorogenic Acid Content In Flos Lonicerae By Hyperspectral Imaging Technology

In recent years,with the deepening of research on Flos Lonicerae,it is not only widely used in medical industry,but also popular in health products,herbal tea,cosmetics industry.It can be seen that the market demand for Flos Lonicerae is also increasing.In order to maintain the stability of Flos Lonicerae industry and people's health,it is necessary to detect and monitor the quality of Flos Lonicerae.Chlorogenic acid,as the most important active component in Flos Lonicerae,has been proved to have significant pharmacological effects on antioxidation,antitumor,antiviral and hypoglycemic.Chlorogenic acid is easy to be affected by temperature and humi dity during the storage of Flos Lonicerae,which makes it degrade,and leads to the decline of the quality,efficacy and safety of Flos Lonicerae.It will not be conducive to ensuring the safety and effectiveness of products with Flos Lonicerae as raw material.The content of chlorogenic acid is used as an important standard to evaluate the quality of Flos Lonicerae,which is of great significance to the detection of chlorogenic acid content in Flos Lonicerae during storage.The traditional detection method of chlorogenic acid content is chemical analysis.Although the method is accurate,it is destructive and time-consuming.Encouragingly,hyperspectral imaging technology has been widely used in food industry with the advantages of fast,nondestructive and accurate.It combines the advantages of spectral technology and imaging technology,and has the characteristics of high resolution,integration of spectrum and multi band.Therefore,Flos Lonicerae with different storage time was taken as the research object,and hyperspectral imaging technology was used for nondestructive detection of chlorogenic acid content in Flos Lonicerae.The correlation between the spectral data of Flos Lonicerae and the content of chlorogenic acid was fully analyzed,the feasibility of hyperspectral technology was discussed,the best spectral pretreatment method was determined,the characteristic wavelengths were selected,the corresponding prediction model was established,so as to improve the reference value for online monitoring the quality change of Flos Lonicerae.The main research contents and conclusions are as follows:1.The feasibility of non-destructive detection of chlorogenic acid content by hyperspectral imaging combined with chemometrics was discussed.Firstly,the whole raw spectrum data of region of interest(ROI)of hyperspectral image was extracted,and the partial least square regression(PLSR),BP neural network and least squares support vector machine(LS-SVM)models were established with the measured of chlorogenic acid content values.The results showed that the determination coefficients(R~2)of PLSR,BP and LS-SVM were all greater than 0.9,which indicated that it was feasible to use hyperspectral imaging technology to detect chlorogenic acid in Flos Lonicerae.2.The effects of different spectral pretreatment methods on the detection results of chlorogenic acid content were compared.The preprocessing methods of standard normal variable(SNV),multiplicative scatter correction(MSC),orthogonal signal correction(OSC),baseline correction(BC),moving average(MA)and derivative were used to preprocess the original spectral data,and PLSR,BP neural network and LS-SVM models were established based on the preprocessed spectral data.Compared with the original data model,the performance of the model could be improved by using the appropriate spectral preprocessing method.The model based on SNV pretreatment had the best performance.The R~2 and root mean square error(RMSE)of prediction set in PLSR model were 0.9766 and 2.711 mg/g,respectively.The R~2 and RMSE of prediction set in BP neural network model were 0.9771 and 2.581 mg/g,respectively.The R~2 and RMSE of prediction set in LS-SVM model were 0.9770 and 2.583 mg/g,respectively.It showed that SNV method could effectively eliminate the spectral errors caused by solid particle size,surface scattering and optical path change,thus improving the accuracy of the model.Therefore,SNV was the best spectral pretreatment method,and the spectral data after SNV pretreatment will be used for further study and analysis.3.The influence of different characteristic wavelength screening methods on the accuracy of prediction model was analyzed,and the best combination of characteristic wavelength screening method and prediction model were determined.To solve the problem of unrelated variables,several single variable selection algorithms,namely,uninformative variable elimination(UVE),competitive adaptive reweighted sampling(CARS),successive projections algorithm(SPA),and four combinations of different variable algorithms(UVE-CARS,UVE-SPA,CARS-SPA,UVE-CARS-SPA),were applied to select characteristic wavelengths from hyperspectral data.The numbers of selected variables were determined as 192,51,17,26,9,12 and 7 by UVE,CARS,SPA,UVE-CARS,UVE-SPA,CARS-SPA and UVE-CARS-SPA methods,respectively.It could be seen from the above results that the spectral dimension of hyperspectral data was reduced,and the input variables of the model could be simplified by using characteristic wavelength screening methods.The PLSR,BP neural network and LS-SVM models were established based on the feature spectrum respectively.The results displayed that the uninformation variables and redundant information could be effectively eliminated by UVE-CARS method,and important feature bands were selected.And the best prediction model combination was determined as UVE-CARS-LS-SVM,which R~2 and RMSE of calibration set were 0.9887 and 1.810 mg/g respectively,and R~2 and RMSE were 0.9785 and 2.496 mg/g,respectively.The results of this study showed that the combination of hyperspectral imaging technology and chemometrics could rapidly,nondestructively and accurately detect the content of chlorogenic acid in Flos Lonicerae during storage,which improved a scientific theoretical basis for effectively monitoring the quality of Flos Lonicerae.