Research On The Detection Models Of Apple Brix And Acidity Based On Visible And Near-infrared Spectroscopy

The internal qualities of apple affect consumers’ purchase intention,and the ratio of brix and acidity directly determines the taste of fruit flesh,so it is particularly important to measure the brix and acidity of apple.Traditional methods of measuring brix and acidity based on physical and chemical analysis are time-consuming and laborious,so how to achieve quantitative and nondestructive detection of apple brix,acidity and other internal quality indexes through an economical and efficient method has become a research hotspot in this field in recent years.With the development of near-infrared(NIR)spectroscopy instruments,Visible-NIR spectroscopy combined with chemometrics has been widely applied in the field of nondestructive detection of fruit internal qualities.Based on this,Visible-NIR spectroscopy was used in this dissertation to study the detection model of apple brix and acidity.On the basis of extensive investigation of domestic and foreign literature in the early stage,the experimental scheme was designed and the following work was carried out:1)In order to verify the feasibility of the experimental scheme,the method for detecting apple brix and acidity was studied.The characteristic spectral bands reflecting the information of apple brix and acidity were determined by investigation,and then the spectra with the wavelength range of 475-925 nm were selected for subsequent modeling.The experiment of pericarp transmittance was designed,and the influence of apple pericarp on spectral data was studied.The experimental results prove that it is reasonable to use diffuse reflection method to collect spectral data of samples.2)The prediction models of apple brix and acidity based on Visible-NIR spectroscopy were established,and the influence of different methods in each process on model accuracy was studied,thus obtaining the optimal prediction models of apple brix and acidity respectively.Results show that the two models are all established by partial least squares(PLS)method based on competitive adaptive reweighted sampling(CARS),and the determination coefficient(R2),root mean square error of prediction(RMSEP)and relative percent deviation(RPD)are 0.9942,0.0674,13.1851 and 0.9949,0.0090,13.9697,respectively.The models all have high accuracy and robustness.3)In order to solve the problem that the number of wavelengths selected by CARS was too large,a wavelength successive increase method was proposed to optimize the models.The results show that compared with CARS,the number of characteristic wavelengths of brix and acidity selected by this method decreases from 141 and 119 to 27 and 40 respectively.The R2,RMSEP and RPD of the optimized apple brix and acidity prediction models are 0.9867,0.1714,8.6662 and 0.9506,0.0279,4.5007,respectively.It can be seen that the proposed wavelength successive increase method not only reduces the number of modeling variables,but also ensures the prediction accuracy and robustness of the models,which proves the effectiveness and versatility of the method.4)In order to realize simultaneous prediction of brix and acidity,an integrated PLS2 model was established.The proposed wavelength successive increase method was applied to select characteristic wavelengths in order to optimize the integrated model.The R2,RMSEP and RDP of the optimized model for predicting brix and acidity were 0.9108,0.2823,3.3486 and 0.8759,0.0365,2.8386,respectively.However,the accuracy of the integrated model was far from that of the independent prediction models for brix and acidity.Therefore,a wavelength successive deletion method was further proposed.The R2,RMSEP and RDP of the integrated model optimized by this method for the two indexes are 0.9780,0.1401,6.7477 and 0.9695,0.0181,5.7246,respectively.The model accuracy and robustness are greatly improved.Compared with the independent prediction models,the accuracy of the integrated model for predicting acidity is improved,while that for predicting brix is guaranteed.The results show that the characteristic wavelengths selected by the proposed wavelength successive deletion method are both effective in predicting brix and acidity,and the established model can realize high prediction precision for the two indexes simultaneously.