| As more and more people are buying apples,they are no longer limited to the selection of appearance quality such as shape and color,but pay more attention to the internal quality indicators such as taste,sweetness and nutrition.The traditional chemical detecting methods are difficult to meet the demand of massive testing and grading because of the damage to apples,tedious sample preparation,and long test time.The NIR(near-infrared Spectroscopy)non-destructive detecting technology is non-destructive,fast,low-cost,and eco-friendly,which is the future trend of the development of apple quality detecting.Given the above problems,in this study,Fuji apples,as the test objects,the contents of the SSC,soluble sugars,and titratable acids were predicted using the technology of NIR spectroscopy and chemical methods.Furthermore,the software and hardware of a portable non-destructive detector for SSC of Fuji apples were designed.Meanwhile,the portable,fast,and low-cost non-destructive detector was developed to test SSC of Fuji apples SSC.The research contents and conclusions of this paper are as follows:(1)The quality detection and correlation analysis of Fuji apples: The weight,cross diameter,and longitudinal diameter of 120 Fuji apples were measured.The SSC of Fuji apples was determined by the refractometer.The reducing sugars and soluble sugars were tested by 3,5-dinitrosalicylic acid colorimetry.The titratable acid was tested by acid-base titration.The spearman correlation analysis was applied to dispose the relativity among the indexes.The results showed that the physical characteristics could not be used to judge the internal quality.There is a significant positive correlation between SSC and soluble sugar,reducing sugar.The correlation coefficients R are 0.785 and 0.504,respectively,which can be used as an ideal detection index to develop the software and hardware of the portable non-destructive detector.(2)Non-destructive detection for the SSC of Fuji apples: The acquisition platform of diffuse reflectance was constructed by FX2000 spectrometer,which was used to collect the spectroscopy information of apple internal quality.Meanwhile,the SSC of Fuji apples was determined by refractometer.The spectral was preprocessed by Savitzky-Golay convolution smoothing method and first-order derivative method.Then,the characteristic wavelengths were obtained using the principal component analysis by correlation coefficient matrices.Finally,modeling and reliability verification were performed by BP neural network algorithm and multiple nonlinear regression.Consequently,the seven wavelengths with the higher scores of 809 nm,833 nm,843 nm,892 nm,927 nm,967 nm,and 979 nm were selected as the characteristic spectrum for the non-destructive detection of apple SSC.The results showed that the prediction model based on multiple nonlinear regression was more effective with a prediction set R of0.9311 and RMSE of 0.5324.(3)Nondestructive detection of soluble sugars and titratable acids in Fuji apples: The spectral information of the samples was collected.Meanwhile,the contents of soluble sugars and titratable acids in Fuji apples were determined by the 3,5-dinitrosalicylic acid colorimetry and acid-base titration methods.The spectral was preprocessed by Savitzky-Golay convolution smoothing method and first-order derivative method.Then,the characteristic wavelengths were obtained using the extract by continuous projection algorithm.Finally,modeling and reliability verification were performed by BP neural network algorithm and multiple nonlinear regression.Consequently,the 11 wavelengths with the higher scores of 752 nm,775 nm,784 nm,804 nm,809 nm,812 nm,814 nm,849 nm,852 nm,879 nm and 882 nm were selected as the characteristic spectrum for the non-destructive detection of apple soluble sugar.The 14 wavelengths with the higher scores of 634 nm,640 nm,670 nm,676 nm,679 nm,689 nm,699 nm,712 nm,715 nm,718 nm,722 nm,728 nm,735 nm and 767 nm were selected as the characteristic spectrum for the non-destructive detection of apple titratable acid.The results showed that comparing the modeling effects of the two models,the soluble sugar prediction model based on multiple nonlinear regression was more effective,with a test set R of 0.9295 and RMSE of 0.5470.The titratable acid prediction model established by multiple nonlinear regression had a prediction set R of 0.8454 and an RMSE of 0.0214;the titratable acid prediction model established by BP neural network had a test set R of 0.7782 and an RMSE of 0.0279.The prediction effects of both models for titratable acid were not very good,It may be related to the fact that the actual measured values of titratable acid were much more influenced by the external environment,and the extracted characteristic spectrum were disturbed by noise.(4)The design and verification of the hardware and software of portable soluble solid content non-destructive detector on basis of NIR diffuse reflectance spectroscopy: Firstly,the hardware part was designed by combining the detection method of characteristic narrowband filter and narrowband LED light source.Then,the best characteristic wavelength of SSC and the modeling method were implanted into the core processor.The modules of main control,light source and light source circuit,and blue tooth were eventually designed.Secondly,the software part of the portable non-destructive detector of Fuji Apples SSC was developed based on Java language.Then,the interfaces of login,user registration,data acquisition,personal information,and password modification were designed.Finally,the spectral information of Fuji apples was collected by the portable non-destructive detector.Then,the prediction model of SSC on the basis of multiple non-linear regression was constructed,meanwhile,their reliability was also verified.The results showed that the prediction model of SSC has a prediction set R of 0.9333 and RMSE of 0.5881.It indicated that the instrument can meet the practical application requirement of portable and fast non-destructive testing SSC. |
PDF Full Text Request