Rapid Detection Method Of Drying Quality Of Honeysuckle Based On Hyperspectral Imaging Technology

Honeysuckle is a kind of commonly used food and medicine in China.It is rich in various nutrients and has very high medicinal and edible value.Due to the instability of chemical components in fresh honeysuckle,it is easy to cause chemical components to react during processing,resulting in changes in color and quality grade of honeysuckle.Color is an important index to evaluate the quality of honeysuckle.At present,people often use sulfur fumigation to maintain the color of honeysuckle during drying pretreatment,which will not only damage people's economic interests,but also damage people's health and seriously affect the sales of honeysuckle products on the market.Chlorophyll is the main chromogenic substance in the chemical constituents of honeysuckle.Because people can not directly distinguish the quality of honeysuckle with naked eyes,the traditional detection method is time-consuming and laborious,which is not suitable for rapid and accurate determination of the quality of honeysuckle.In this study,hyperspectral imaging technology combined with chemometrics was used to detect and analyze the changes of sulfur residue and chlorophyll content in honeysuckle processing.The research results can provide a theoretical basis for the rapid nondestructive testing of honeysuckle quality by hyperspectral imaging technology,and also provide a reference for the application of hyperspectral imaging technology in the field of non-destructive testing of food processing quality.The main research contents and conclusions are as follows:1.Hyperspectral imaging technology combined with chemometrics was applied to develop a predictive model for detecting sulfur-fumigated honeysuckle samples with four concentration gradients of 0.0%,0.5%,1.0%and 1.5%on a fresh mass basis.Firstly,hyperspectral imaging technology was used to collect spectral image data of honeysuckle moving average(MA),normalize,Savitzky-Golay filter(SG),multivariate scattering correction(MSC)and standard normalized variate(SNV)were used to pretreat the original spectrum data,and SG was the best pretreatment method.Then,the spectral information pretreated by SG was used to establish fisher discriminant analysis(FDA)and kernel fisher discriminant analysis(KFDA)models.The result showed that KFDA model had better discriminant accuracy.Finally,three characteristic extraction methods,regression coefficient(RC),Wilks and RC-Wilks,were compared.The results showed that the discriminant accuracies of the three methods were 100.0%after SG.The KFDA model based on characteristic wavelengths data selected by RC-Wilks criterion could achieve shorter calculation time and better inter-class distribution.2.The quantitative model of chlorophyll content in honeysuckle with different drying times was established by hyperspectral imaging technology.The hyperspectral data were collected randomly from 5 groups of dried honeysuckle after 0,2,4,7 and10 h,and pretreated by MA,SG,MSC and SNV.The best pretreatment method was SG.Then,RC,ssuccessive projections algorithm(SPA)and competitive adaptive reweighted sampling(CARS)were used to select the characteristic variables from the spectral data pretreated by SG,and the partial least square regression(PLSR)and least squares support vector machine(LS-SVM)models were established,respectively.The results showed that CARS-LS-SVM model was the optimal prediction model.The determinant coefficients(R~2)of the training set and testing set were 0.9864 and0.9692,and the root mean square errors(RMSE)were 0.0342 and 0.0519,respectively.3.Image and spectral fusion information was used to detect changes in the chlorophyll content of honeysuckle at different drying times during processing.This study used principal component analysis(PCA)to reduce the dimensionality of the hyperspectral image to obtain three principal component images.Among them,the gray-level co-occurrence matrix(GLCM)algorithm was used to extract the texture features from the obtained principal component image to obtain single texture feature information.The spectral feature wavelengths were extracted by CARS to obtain single spectral feature information.The combination of texture and spectral feature wavelengths was optimized using PCA to obtain the optimized texture and spectral feature combination.Then,based on the single texture feature,single spectral feature,a combination of texture and spectral features and optimized texture and spectral feature combination,the discriminant analysis model of LS-SVM was established.The results showed that the LS-SVM discriminant analysis model established by PCA-optimized texture and spectral feature wavelength combination was the optimal detection model for the chlorophyll content of honeysuckle at different drying time.TheR~2 of the training set and testing set were 0.9917 and 0.9804,and the RMSE were 0.0201,0.0324,respectively.