Optical Nondestructive Analysis Of Citrus And Citri Reticulatae Pericarpium And Software Development

The combination of near-infrared spectroscopy combine with chemometrics was applied for the non-destructive identification and quantitative quality detection of citrus as well as citri reticulatae pericarpium was applied in this paper.The penetration ability of near infrared diffuse reflected light on citrus peel was examined.Different pretreatment and variable screening methods were compared to gain the optimal chemometrics methods.To construct the identification model and quality quantitative analysis model of citrus and citri reticulatae pericarpium,the influence of the measured points was discussed.The main results and conclusions obtained are as follows:1)The penetration ability of near-infrared diffuse light to different thickness of peel and the ability of collecting flesh information were investigated.The penetration ability of near infrared spectroscopy diffuse light to on fruit peel of 'Shatangju' mandarin,mandarin,fertile orange,sweet orange,grapefruit and citrus maxima(Burm.)Merr.cv.Shatian Yu were compared and analyzed,and the spectral data obtained were clustered by principal component analysis(PCA)method.The results showed that the near-infrared diffuse reflection light could penetrate the thin peel of the citrus peel of 'Shatangju' mandarin,mandarin,fertile orange,sweet orange,but near-infrared diffuse light could not be obtained the internal information by the thick peel(thickness over 2 cm)fruits of grapefruit and citrus maxima(Burm.)Merr.cv.Shatian Yu.2)The non-destructive identification analysis of citrus in different origins,different varieties and different-age citri reticulatae pericarpium in different years was achieved by the combination of near infrared diffuse reflectance spectroscopy as well as chemometric methods.The influence of different spectral collection points of citrus and citri reticulatae pericarpiumwas investigated.The single and combined pretreatment methods were used to eliminate the background and noise interference in the spectrum.The PCA and linear discriminant analysis(LDA)were used to realize the identification analysis of citrus in different production areas,varieties and different-age citri reticulatae pericarpium.The results show that the linear discriminant analysis method combined with the best sampling method can achieve 100% discriminant analysis for the analysis of the above samples.3)The quantitative analysis of soluble solids contentin 'Bingtang' sweet orange was realized through the combination of the spectroscopy based on the portable visible near infrared and chemometrics.The single and combined pretreatment methods were used to eliminate the background and baseline drift interference in the spectrum of 'Bingtang' sweet orange.A robust model of soluble solid content in 'Bingtang' sweet orange was established by combining the methods of variable selection and modeling.The results show that the standard normal variable transformation(SNV)method is the best pretreatment method for near infrared spectrum of 'Bingtang' sweet orange,and the accurate quantitative model is obtained by combining the variable adaptive boosting partial least squares(VABPLS).The prediction correlation coefficient of the model is 0.824,and the root mean square error of the prediction set is 0.579.4)The quantitative analysis of soluble solids contentin Hunan fertile orange was realized through the combination of the near infrared diffuse reflectance spectroscopy and chemometric methods.Different pretreatment methods were used to eliminate the interference factors in the spectrum of fertile orange.In this paper,different methods of variable selection were introduced to screen the spectral data variables of Hunan fertile orange.These variable selection methods and different spectral acquisition methods help to construct the most robust quantitative model of soluble solid content in Hunan fertile orange.The results show that,monte carto uninformative elimination(MC-UVE),the best method for variable selection,combined with the average spectral data of 4 equatorial lines can construct a reliable and robust quantitative model for the content of soluble solids contentin Hunan fertile orange.The prediction correlation coefficient of the model could reach 0.854,and the root mean square error of the prediction set is 0.715.5)Based on MATLAB software,the softwares of spectral pretreatment and VABPLS for the quantitative analysis were developed.