Variety Discrimination And Non-destructive Detection Of Quality Of Pomegranate By Near-infrared Spectroscopy

Pomegranate has 2000 years of cultivation history in China,and the current planted area of about 175 million mu,ranking first in the world.However,during the process of pomegranate cultivation and storage,only relying on manpower could not meet the standardization requirements because of the wide variety of pomegranate and lacking of the effective and rapid monitoring methods.Near-infrared spectroscopy has many advantages,for example,simple,fast and no pollutation;so it has been used in the qualitative identification of varieties and the quantitative detection of quality.At present,research objects of near infrared technology mainly are the thin-skin fruits,and for thick-skin fruits such as pomegranate fruit are few.In this study,the near-infrared prediction model of varieties and quality of pomegranates in Shaanxi lintong were established by using diffuse reflectance spectroscopy and chemometrics.The research provided software support for non-destructive detection techniques and classification for pomegranates,promoting the rapid development of pomegranates industry.The main contents and results were as follows:(1)In order to identify varieties of pomegranates in Shannxi lintong,the near infrared spectrum of three different varieties of pomegranates were collected and processed by MSC.Models were established using the principal component combined with Multilayer Perceptron neural network(PCA-MLP),the principal component combined with Fisher discriminant analysis(PCA-Fisher-DA),Partial least square discrimination analysis(PLS-DA).By comparing the three discriminant models,discriminant result of PLS-DA was superior to others.The correct recognition rates for validation sets were 97.30%,96.55% and 96.77%,respectively in the full spectrum.In order to further improve the discriminant rate of PLS-DA model,the characteristic bands were selected by "loading method",and the correct recognition rates of the model after optimization were 97.30%,100.00% and 100.00%.It was shown that optimized PLS-DA model can achieve the identification of pomegranate varieties.(2)In order to reduce the effect of variety on the quality parameters,three different varieties of pomegranates were selected as the research objects.The PLS method was used to establish local variety and hybrid variety models for pH value of pomegranate.After comparison,it was found that the model established by mixing the three cultivars had a betterprediction effect.The correlation coefficients of the calibration set and the verification set are all greater than 0.900.Therefore,the model established by the calibration set of three mixed varieties can achieve rapid and accurate determination for pH in pomegranates.This conclusion could further extended to the determination of other quality parameters.Comepetitive adaptive erweighted sampling was implemented to select effective variables of NIR spectroscopy of quality parameters of pomegranate.The R~2 of the validation set of TA,SSC,SSC/TA were 0.903,0.930 and 0.853,respectively.The RMSEP were 0.019%,0.244 °Brix,2.142,respectively.(3)In order to establish PLS models of polyphenols and flavonoids in pomegranate seeds,the near-infrared spectra of whole pomegranates were collected.The correlation coefficient between the calibration set and the verification set of the established PLS model is less than 0.850.The prediction accuracy of the models was poor,and was needed for further study.For established PLS model of anthocyanins,The R~2 of the calibration set was 0.881,RMSEC was 1.318 mg/100 g,the R~2 of prediction set was 0.863,the RMSEP was 1.266mg/100 g.