| Pear is one of the oldest plants cultivated by human beings,and its variety and fresh fruit quality have been the main concerns for consumers when purchasing.Among the quality indicators of pears,soluble solids content（SSC）and titratable acid（TA）are two important factors affecting the quality of pears.The surface color plays an important role in the process of fruit grading and consumer purchase.In this study,to address the problems of large differences in pear fruit yield,different fruit quality after harvest,and better fruit appearance quality but poor edible taste,the feasibility of rapid quantitative and qualitative analysis of pear fruit based on Fourier transform near infrared spectroscopy（FT-NIR）was carried out,and a model for non-destructive testing of pear fruit quality and variety identification was established to provide a reference value for production practice.The main study components and findings are as follows:（1）The preliminary modeling process for‘Zaosu’and‘Dangshansu’was based on the spectral data of 340 fresh pear fruits,and the quantitative calibration model was established by partial least squares（PLS）combined with cross-validation.The results showed that the best results were obtained for the calibration models of SSC and TA of‘Dangshansu’using FT-NIR（RPD:3.272 and 2.239）,which were better than those established for SSC and TA of‘Zaosu’（RPD:1.407 and 1.471）.（2）Based on the initial calibration model,the parameters of the model were supplemented and validated in the following year.When conducting modeling,‘Zaosu’and‘Dangshansu’pears were selected by choosing different wavelength ranges and spectral pre-processing methods to select the best parameters for modeling SSC,TA,color value L*,color value a*,and color value b*.The quantitative NIR PLS models of SSC,TA,color value L*,color value a*,and color value b*were developed under optimal conditions.The results showed that the models developed using FT-NIR had the best SSC models for‘Zaosu’and‘Dangshansu’pears（RPD:2.56 and 2.28）,and the‘Zaosu’color value a*and‘Dangshansu’pear TA,color value L*,color value a*models are better（RPD:1.67,1.87,1.75,1.66,and 1.81）.The good predictive ability of the models also further demonstrated the operational feasibility of using NIR spectroscopy for the fruit quality of‘Zaosu’and‘Dangshansu’.（3）In order to prove that the model has better generalizability,PLS modeling was performed for‘Yuluxiang’and‘Koral’pear under model-optimal conditions.The RPD of the SSC,TA,and color value b*of‘Yuluxiang’and the SSC,color value L*,color value a*,and color value b*calibration models of‘Koral’pear ranged from 1.4 to 2.27,and the RPD of the corresponding prediction sets ranged from 1.67 to 2.44,indicating that the established NIR model has good prediction ability.In conclusion,FT-NIR spectroscopy is an effective tool for rapid,non-destructive analysis of pear fresh fruit quality.（4）In the identification of pear varieties based on NIR spectroscopy,the pre-processing method of spectral data and the selection method of the feature band are very important.The second-order derivative（SD）preprocessing of the spectra in the wavelength range of7500-5000 cm-1 and the use of the first-range calibration method and the reproduction level normalization method were able to classify‘Zaosu’and‘Dangshansu’accurately.This indicates that FT-NIR spectroscopy can be used for variety identification of‘Zaosu’and‘Dangshansu’. |
