| Jinyu No.1 peach,also known as red is not soft.Ripe Jinyu No.1 peach is attractive in color,rich in taste and high in VC content,which is one of the fruits that people buy more frequently in summer.The variety is a bud-variety variety,which has strong resistance to stress,is resistant to infertility,and has no special pests and diseases.Therefore,it is widely planted in different altitudes.In addition to the advantages of planting,the variety is colored earlier,and the red color is hanged in early July;the fruit has a hard texture and can be stored for about 20 days at room temperature;the market has a long sales period and can be extended from mid-August to early September.This paper focuses on the Jinyu No.1 peach in different producing areas in Shanxi Province,and conducts an experimental study on its internal quality.Main research contents and main conclusions:(1)The LB10 T hand-held sugar meter with range 0 to 80% was used to measure the soluble solid content of the 100 Jinyu No.1 peach samples picked from Taigu County.Amount to 200 SSC values on the sunlit side and the shade side were analyzed statistically.It was found that the maximum,minimum,and mean values of the soluble solids concentration on the sunlit side 15.47°Brix,9.2°Brix,and 11.25°Brix respectively,was higher than that in the shade side 13.07°Brix、8.5°Brix、10.68°Brix respectively.And among them,the average SSC of the sunlit side of 84 samples was higher than that of the shady surface.(2)Using the Monte Carlo sampling(MCCV)outlier detection method,the abnormal were excluded from samples 100 samples of Jinyu No.1(including 200 spectra).The experiment set the mean threshold to6,and the variance threshold of 1.5,the value out of the threshold range will be removed.The result No.1,No.18,No.60,No.97,No.98,No.99,No.100,No.110,No.111,and No.112 were determined as abnormal value and excluded from the spectral matrix.(3)Several common spectral pre-processing methods were compared in the experiment,including Multivariate scattering correction(MSC),Savitzkye-Golay Smothing,and Standard normal variatetransformation(SNV),First derivative and second derivative(1-Der and 2-Der),median correction(MF).And PLS algorithm was utilized to evaluate the performance of different spectral pre-processing methods.The results showed that the Savitzkye-Golay(3 points)smoothing pre-treatment has the best effect.The correction set correlation coefficient(Rc)and root mean square error(RMSEC)were 0.9147 and 1.6615,respectively;the prediction set correlation coefficient(Rp)and root mean square error are(RMSEP)were0.8891 and 1.5689 respectively.(4)Characteristic wavelength extracted by competitive adaptive weighting algorithm(CARS)and the random frog(RF)algorithm and the full-band spectrum were used to establish the local position model of PLS,the average spectral model and the global position model for the peach sample,respectively.The SSC value was quantitatively predicted.The results showed that the prediction accuracy of PLS model based on characteristic wavelength was higher than that of full-band modeling.The RF-PLS local position model had the best performance.The Rc of the sunlit side and shady side were 0.9683 and 0.9566,respectively.The Rp and RMSEP values of the prediction set were 0.9564 and 0.7043 on the sunlit side and 0.9528,0.7965 on the shady side.The characteristics spectral of all samples were classified by 500-750 nm and1400-2200 nm spectral data.The SPA-PLS and SPA-LS-SVM discriminant models were established respectively.The optimal classification model was built based on the characteristics spectra from500-750 nm band.Overall discriminant accuracy of the model based on the SPA-LS-SVM model was96.11%.To combine the above-mentioned local position model and the classification models of surface feature effect.In order to improve the prediction accuracy of SSC from Jinyu No.1 Peach,30 unknown Taigu County Jinyu No.1 peach samples were compiled as independent verification sets to establish the peach sample SSC predictive compensation model.A satisfactory result was obtained.The correlation coefficient of the compensation model was 0.9532,and the predicted root mean square error was 0.7256°Brix.(5)Jinci No.1 peach from multiple producing areas to establish hybrid model with strong inclusiveness and improve the prediction accuracy of SSC.After the spectral data pre-processed by SG+MSC,Monte Carlo elimination of non-information variables combined with successive projection algorithm were used to reduce the dimension of the spectra data.Finally,17 variables were selected.The calibration sets from single origin,two mixed and three mixed areas were used as modeling inputs to establish the PLS models and ELM models.It is found that the prediction accuracy of the PLS model basedon the mixed three region correction set achieved the best result with Rp of 0.949 and RMSEP of 0.652°... |
PDF Full Text Request