Research On ’Lingwuchang-zao’ Jujube Internal Components Non-destructive Detection Methods Based On Hyperspectral Technology And Hyperspectral Fluorescence Technology

’Lingwuchangzao’-jujube is a special agricultural products and has advantage in Ningxia.Therefore,it is important to explore non-destructive detection techniques for the quality of’LingWuchangzao’-jujube effectively whose result can be used to provide a theoretical reference including the industrialization,serialization and high value-added of the jujube.This thesis will regard the core quality of the jujube as the testing index,such as Vitamin C,Oxalic Acid,Malic Acid.The nutrient content of Oxalic Acid,Malic Acid and Vitamin C in jujube is measured by HPLC.The self-building hyperspectral fluorescence system,Using markov distance,leverage value,studentized residuals and cook distance to eliminate the abnormal samples.The original spectrum were pretreated by SVM,moving average smoothing and normalize method,Using iPLS and GA methods to extract the characteristic wavelength,Using PCR,SVM and PLS methods to establish the prediction model of the internal components of the jujube.By combining theory calculation of the various functional groups and characteristics of molecular structure with the absorption peak to explore the absorption mechanism of jujube in near infrared spectrum.The primary conclusions are as follows:(1)This thesis researches the content of the oxalic acid,malic acid and vitamin C with various durations of storage.In the early stage of the storage,there is polysaccharides and uronic acid synthesized into vitamin C whose contents are increased,but declined slowly at the late stage.During the storage,the content of oxalic acid increased slowly.Moreover,in the late storage,Vitamin C will convert to Oxalic Acid.While the content of Malic acid remains invariable.(2)Research on ’lingwuchang-zao’jujube non-destructive detection based on near infrared hyperspectral technology.The results showed that the effect has a high correlation for vitamin C original spectrum pretreated by GF,oxalic acid original spectrum pretreated by the average smoothing and malic acid original spectrum using the PLS model.The modeling spectral was chosen 5,7,10,14 interval to establish model between the spectral and the Vitamin C content by means of ipls method,the modeling spectral was chosen 11,12,13 interval to establish model between the spectral and the oxalic acid content by means of ipls method,the modeling spectral was chosen 10,11,13,14,15 interval to establish model between the spectral and the malic acid content by means of ipls method.Using GA method to choose the characteristic wavelength,while vitamin C spectrum selected 34 bands,oxalic acid spectrum selected 33 bands,malic acid spectrum selected 54 bands,the characteristics wavelength which was selected by iPLS and GA method contains much information of whole-band.But other band also has contribution to the model,so the characteristic wavelength of the model have a lower correlation than whole-band model.(3)Study on mechanism of near infrared spectral absorption of ’lingwuchangzao’-jujube components.Vitamin C,oxalic acid and malic acid in 900-1700nm band have 2,4,3 characteristic absorption peaks,which are corresponding to the theoretical calculation of different functional groups.The characteristic absorption peaks is corresponding to the characteristic optimization wave band of chose by iPLS and GA method whose result can be used to interpret the mechanism of near infraredhyperspectral model.(4)Research on ’lingwuchang-zao’jujube non-destructive detection methods based on the hyperspectral fluorescence.The results show that the SVM method of jujube vitamin C content after pretreating the baseline is optimal.Rp=0.84,jujube oxalic acid content PLS model is optimal after de-trending pretreated.Rp= 0.86;jujube malic acid content PLS model is optimal after MF pretreated.Rp=0.84.By comparing the hyperspectrum fluorescence prediction model and the near infrared hyperspectrum prediction model,the result show that the hyperspectrum fluorescence PLS prediction model of correlation coefficient is significantly higher than the near infrared hyperspectral PLS prediction model of correlation coefficient.