Study On The Method Of Detecting The Origin,Shelf Life And Sugar Content Of Orange Based On Hyperspectral Imaging Technology

Nowadays,the quality of life is constantly improving,and people prefer healthy and safe fruit and vegetable products.In the 1990 s,the total output of tangerines ranked first among all kinds of fruits in the world.Because tangerine contains a variety of nutrients,the content of vitamin C is 6 to 20 times that of apple,and it is one of people’s favorite fruits because of its nutritional and medicinal value.If the surface of tangerine is damaged and the internal quality changes,these quality and safety problems will seriously affect the sales and processing of tangerine and will seriously affect the health of consumers.Based on hyperspectral technology and texture feature fusion,the origin,shelf life and sugar content detection of high-quality tangerine can be distinguished,which ensures high-quality tangerine products.Main research contents and conclusions:(1)Origin identification of orange based on hyperspectral technology.Four kinds of samples(Dongjianghu Orange,Xunwu Orange,Yaowan Orange,Yongquan Orange)were collected,100 samples for each variety.Spectral analysis of oranges from different habitats,principal component analysis PCA of oranges from different habitats,model discrimination analysis based on full band,pretreatment RAW,MSC,Baseline,DF,and characteristic wavelength screening to establish PLS-DA model and SVM-DA model.Comparing the accuracy of the prediction set of the model,it is concluded that SVM-DA model is superior to PLS-DA model in all models,and the accuracy of the prediction set is the highest.UVE and CARS algorithms are used to screen the characteristic wavelengths of the original spectrum and the pretreated spectrum,and the SVM-DA model is established again for the filtered wavelengths.According to the discrimination result of the SVM-DA model established based on the characteristic wavelength,the correct rate of the confusion matrix prediction set of the prediction results of MSC-UVE-SVM-DA and MSC-CARS-SVM-DA models is the highest,and the identification of orange origin shows great advantages.(2)Study on the shelf-life discrimination of citrus based on the fusion of spectral and texture features.There are 100 experimental samples,the variety is Nanfeng orange,which are placed on the 0,3,6 and 9 days respectively.Spectral characteristics of different shelf life were analyzed.PCA of spectral characteristics of different shelf life.In this study,PCA is used to reduce the dimensions of all sample images,and the GLCM of sample feature images is used to extract texture features.In this study,UVE and CARS algorithms are used to establish PLS-DA model for spectral characteristics respectively.The PLS-DA model based on wavelength screening has a high accuracy.(3)Study on the optimal detection location of sugar content of Yongquan orange based on hyperspectral imaging technology.In this experiment,120 Yongquan mandarin oranges were used as experimental samples to establish a prediction model for the correction set of the calyx,equator,fruit stem and global position of the mandarin oranges.Comparing the full variable models of different parts,Raw PLSR model and Baseline LSSVM model have better prediction performance for calyx parts;Baseline PLSR model and MSC-LSSVM model had better prediction performance for fruit stem parts;For equatorial region,SNV-PLSR model and SNV-LSSVM model have better prediction performance;For the global,MSC-PLSR model and MSC-LSSVM model have good prediction performance.Comparing different models based on characteristic wavelengths,no matter which model is,it is the best in the whole world;For the global,MSC-CARS-LSSVM model is better than MSC-CARS-PLSSR model.Therefore,MSC-CARS-LSSVM model based on global spectral data is the best for sugar content prediction.The research in this paper shows that the research on the origin,shelf life and sugar content detection methods of oranges based on hyperspectral imaging technology can provide technical support and reference for the internal and external quality detection of fruits and can also provide reference for the sales and processing of fruits.