Research On Identification Of Lycium Barbarum Variety And Detetion Of Fumigated And Dyed Lycium Barbarum Based On Hyperspectral Imaging Technology

Lycium barbarum is widely grown in Northwest China.It is not only a food raw material but also a medicinal material with health care,antioxidant,and other medicinal effects.Lycium barbarum is rich in polysaccharides,flavonoids,amino acids,and phenolic substances,and there are significant differences in the content of active chemical components in different varieties of Lycium barbarum.In the sales process,some unscrupulous merchants not only use shoddy products but even fumigate and dye the aged wolfberry by using sulfur and Sudan red to achieve the effect of bright color,which seriously endangers the health of consumers.Traditional identification methods for wolfberry include chemical titration,near-infrared spectroscopy,etc.However,there are many disadvantages to these methods.Collecting samples through chemical titration would damage them and take a long time.The spectral data collected by point sampling of infrared spectroscopy is difficult to characterize the whole sample.The hyperspectral imaging technology can collect comprehensive spectral information of the whole sample,and achieve good detection effectiveness in non-destructive testing of crops.So based on the above analysis,the non-destructive testing of different varieties of Ningxia Lycium barbarum and smoked Lycium barbarum is studied by using hyperspectral imaging technology combined with machine learning algorithms in this paper.And the main research contents and conclusions of this paper are shown as follows:(1)Collect spectral data and study the detection mechanism.First,the hyperspectral image acquisition system was used to collect the hyperspectral images of 6 different varieties of Lycium barbarum and Lycium barbarum with sulfur fumigation and Sudan red dyeing.And the spectral data was extracted with a single Lycium barbarum as the region of interest(ROI)by using the threshold segmentation method.The spectral curve was then analyzed in conjunction with changes in the content of available chemical substances inside,which verified the feasibility of identifying different varieties of Lycium barbarum,fumigated and dyed Lycium barbarum by hyperspectral imaging technology.IV(2)Identification of Lycium barbarum variety based on hyperspectral imaging technology.After extracting the spectral data,the variational mode decomposition(VMD)was introduced to denoise the spectral data with the modal parameter k set to different values in the data preprocessing process.And support vector machine(SVM)model was established to compare and analyze the pretreatment effects under different modal parameters and compared with the multiplicative scatter correction(MSC),standard normal variate(SNV),and Savitzky-Golay smoothing(SG).Eventually,it was found that VMD can play a prominent role in preprocessing the original spectral data of Lycium barbarum.Then,competitive adaptive reweighting sampling(CARS),iteratively retains informative variables(IRIV),and variable iterative spaec shrinkage approach(VISSA)were used to screen the preprocessed data.Besides,the SVM model was established,and the screening effects of different feature selection methods were compared and analyzed through the model classification accuracy.Finally,the WOA and the SMA were used to optimize the penalty factor c and the kernel function parameter g of the SVM model,and made a comparative analysis as well.The results showed that the accuracy of the training set and prediction set based on the VMD-CARS-WOA-SVM model could reach 96.7% and 94.2%.This improved both the classification accuracy.(3)Detection of fumigated and dyed Lycium barbarum based on hyperspectral imaging technology.In this study,aged Lycium barbarum was fumigated and dyed with sulfur and Sudan red respectively,and then samples of fumigated,dyed,and fresh Lycium barbarum of similar size and color are used for experiments.After extracting the spectral data,we set the modal parameter k in VMD to 5.Next,the original spectral data of smoked Lycium barbarum was preprocessed,and then the result of it was compared with other preprocessing methods.It was found that VMD could also play a role in preprocessing the smoked Lycium barbarum spectral data.Then,CARS,IRIV,and VISSA were used to screen the preprocessed data,and the result showed that the feature selection effect of VISSA is the best by establishing the SVM model.Finally,the SVM penalty factor c and kernel function parameter g were optimized using WOA and SMA.It followed that the accuracy of the training set based on the VMD-VISSA-SMA-SVM model reached 98.1%,and the accuracy of the prediction set reached 96.7%.Compared with the original SVM model,the classification accuracy of the VMD-VISSA-SMA-SVM model was greatly improved.(4)The Lycium barbarum variety,fumigated and dyed Lycium barbarum detection system was designed,and users could choose corresponding functions according to their needs.The system interface included interfaces of a function selection,data import,data preprocessing,feature band screening,and detection results.After that,the detection accuracy of the system was tested by importing new sample data.The results indicated that the identification accuracy of Lycium barbarum variety reached 95%,and the detection accuracy of fumigated and dyed Lycium barbarum reached 93.3%.