Spectral And Imaging Detection For Assessing Quality Of Fresh Jujube Under The Influence Of Multiple Source Factors

Jujubes are rich in various nutrients,and has high edible and medicinal value.The detection model is not ideal in accuracy,processing speed is slow,the reliability and stability of the algorithm are not high,and the detection models built by different instruments are difficult to share,these problems are common in fruit quality testing because of the influence of growth environment,maturity and tree age.Huping jujube was used as research object in the study.In order to improve the accuracy and stability of detection and realize synchronous detection of multiple indicators,visible/near infrared spectroscopy,near infrared hyperspectral imaging,chemometrics and data mining technology were used in this study to detectsoluble solids content(SSC)and vitamin c content(VC).The main contents and conclusions of this study were as follows:(1)To perform accurate and synchronous detection of the SSC in fresh jujubes at different stages of maturity,hyperspectral imaging was used to establish robust models.The combined data constituting four maturation stages were used to build the grid-search least squares support vector machine(GS-LS-SVM)model.The determination coefficient(Rp2),the root mean-square error(RMSEP),and the residual predictive deviation(RPD)of the prediction set for samples of the overall stages were 0.98,1.10%,and7.85,respectively.Furthermore,a successive projections algorithm(SPA)was used to extract the characteristic wavelengths of the combined data.An artificial bee colony(ABC)algorithm(for the prediction set,Rp2= 0.98,RMSEP = 1.19%,RPD = 7.25)was used to improve the SPA-LS-SVM model,which was better than the SPA-GS-LS-SVM model(for the prediction set,Rp2=0.98,RMSEP=1.24%,RPD=6.96).Lastly,visualization of the SSC distribution map was performed based on the SPA-ABC-LS-SVM model,which clearly showed that the SSC gradually increased during maturation.The results indicated that it was realistic to construct a detection model of the multimaturity stage.This research also demonstrated that the combination of hyperspectral imaging and the ABC had good application values in the testing of agricultural products.(2)To solve the difficulty of sharing detection models built by different instruments,the transfermethod of the soluble solids content(SSC)detection model between instruments were explored using visible/near infrared spectroscopy.SSC detection models were established by least squares-support vector machines(LS-SVM),respectively.The Shenk's,direct standardization(DS)and slope/bias(S/B)algorithm were used for model transfer,respectively.Then,according to the regression coefficient,the sensitive wavelengths of the master instrument(24)and the slave instrument(28)were extracted.24 single variables(SV),23 common variables(CV)and 29 fusion variables(FV)were selected,all of which contained the main absorption bands of SSC.LS-SVM detection models of the master instrument were respectively established by the preferred variables,which(Rp2=0.78~0.80,RMSEP=1.07%~1.13%)was better than the model built by the full wavelength(Rp2=0.73,RMSEP=1.36%)for the prediction result of the master instrument.However,the model failed in predicting spectra from different instruments(RMSEP=6.62%~7.88%).(3)Based on the wavelength position shift and the absorbed property of molecular vibration,these algorithms named as common variable-subtraction correction(CV-MC),single variable-subtraction correction,fusion variable-subtraction correction and common variable-wavelength correction(CV-WC)were respectively proposed for model transfer.These methods were compared with SV-Shenk's,CV-Shenk's,FV-Shenk 's,SV-DS,CV-DS,FV-DS,SV-S/B,CV-S/B and FV-S/B algorithms.The results showed that the prediction results(Rp2=0.03~0.34,RMSEP=2.44%~4.67%)were poor when the model was transferred by the full-band.Using the model built by the preferred variables,the results transferred by SV-Shenk 's,CV-Shenk 's and FV-Shenk's were poor,and the results transferred by other algorithms(Rp2=0.47~0.73,RMSEP=1.30%~1.90%)were better than the full wavelength.The CV get better transfer results than the SV and the FV,and the CV-MC result was the best(Rp2=0.73,RMSEP=1.30%).The predicted result after CV-WC transfer(RMSEP=1.62%)was similar to CV-DS and CV-S/B.The research indicated that both CV-MC and CV-WC were effective model transfer algorithms,which were of great significance for establishing a common jujube quality detection model between different instruments.(4)To investigate the impact of cultivation mode on detection models,the SSC detection model was established in the range of 450~780nm(Vis),780~1100nm(SW-NIR),1100~2400nm(LW-NIR),780~2400nm(NIR),and 450~2400nm(Vis/NIR),respectively.The prediction results of the models built in the Vis and LW-NIR bands were not ideal.The models built in SW-NIR?NIR and Vis/NIR range all had good prediction results for the same conditional jujubes.When predicting jujubes with other conditions,themodel built by the open field jujubes at 450~2400nm predicted the results of rain shelter verification set jujubes(Rv2=0.73,RMSEV=1.51%,RPD=1.93),which was better than the results of other models(Rv2=0.42~ 0.74,RMSEV=1.74%~2.70%,RPD= 1.01~1.57).RC,CARS,F-CARS,R-CARS,RF and SPA algorithms were used to extract characteristic wavelengths,the RC-PLSR model had the best results(Rv2=0.83,RMSEV=1.17%,RPD=2.33)when predicting the open field samples.When predicting rain shelter jujubes,the F-CARS-PLSR model had better prediction results(Rv2=0.71,RMSEV=1.70%,RPD=1.84)than the models established by other five methods(Rv2=0.64~0.71,RMSEV=1.79%~2.11%,RPD=1.38~1.63),and the prediction results were not ideal.This study showed that cultivation mode had a certain impact on the SSC detection model.(5)Based on characteristic wavelength extracted by RC,the model updating method with added samples by KS algorithm was used to optimize the synchronous detection model of open and rain shelter jujubes.The Rv2 of rain shelter verification set jujubes was increased to 0.8 from 0.7 before updating,the RMSEV of rain shelter verification set jujubes was decreased to 1.32% from 1.85% before updating,and the RPD was increased to 2.21 from 1.58 before updating,which was better than the verification results after full-wavelength updating(for the result of rain shelter verification set jujubes,Rv2=0.79,RMSEV=1.36%,RPD=2.15).(6)To investigate the impact of tree age on detection models,a SSC detection model for different tree age samples under different cultivating conditions was established by PLS.The model built by a single tree age had poor prediction ability for other tree age jujubes,and the model built by a single tree age under a single cultivating condition had poorer prediction results for other tree age jujubes under other cultivating conditions than for other tree age jujubes under the same cultivating condition.The model built by combining the two tree ages had slightly better prediction results for other tree age jujubes than the model built by a single tree age.The model built by combining the three tree ages had slightly better prediction results for different tree age jujubes(Rp2=0.70~0.78,RMSEP=1.11%~1.42%,Rv2=0.76,RMSEV=1.40%)than the model built by a single tree age and combining the two tree ages,and implemented predictions for different age jujubes.This research revealed that tree age was a factor that affects the performance of the calibration model.(7)To achieve synchronous detection of SSC and VC for fresh jujube with different cultivation conditions and different ages,RC,SPA,F-CARS,R-CARS,R-CARS-SPA,and F-CARS-SPA methodswere used to extract characteristic wavelengths of SSC and VC,respectively.PLS and LS-SVM detection models are established separately,respectively.And the selection and fusion of characteristic wavelengths of the SSC and VC was performed.Since there were the same and similar wavelengths between SSC and VC characteristic wavelengths,new characteristic wavelengths were extracted by eliminating wavelengths that contain the same information in combination with wavelengths position offset.Common wavelengths and fusion wavelengths were selected again among the characteristic wavelengths fused by different methods.Then,the LS-SVM was used to build the detection model.The results showed that the fusion wavelength moderately increased the effective information of the sample,improved the prediction performance of detection to a certain extent,and the prediction result(For the SSC of the prediction set,the Rp2 was 0.78 and the RMSEP was 1.17%.For the VC of the prediction set,the Rp2 was 0.80 and the RMSEP was 0.29 mg.g-1.The Rv2 and RMSEV for the SSC of the independent verification set were 0.80 and 1.29%,respectively.The Rv2 and RMSEV for the VC of the independent verification set were 0.81 and0.27 mg.g-1,respectively.)was better than the results of the model built by other methods.This study realized the synchronous detection of SSC and VC,and provided some theoretical and method support for the realization of the simultaneous detection of multiple indicators in fresh jujube under different conditions and the improvement of the universal applicability of detection model.