Research On Non-destructive Testing Of Pear Fruit Quality Based On Hyperspectral Technology

In recent years,with the improvement of people’s living standard,people’s demand for fruit is also increasing,which occupies a large part of people’s life.The pear is one of the daily favorite fruits.Besides the appearance of pear,people pay more attention to the inner quality and nutrition of pear.Therefore,the requirements for pear quality inspection are higher.At present,the methods of fruit quality detection are time-consuming,laborious,with large error and destructive to fruits.So we need a method that can detect fruit quality quickly and accurately and nondestructively.Hyperspectral technology is a popular technique in nondestructive testing of fruit quality in recent years.However,there is little research on pear fruit nondestructive testing.In this paper,huangguan pear and dangshansu pear were used as experimental materials,and the contents of soluble solids,hardness and water in pear fruits of the two varieties were studied by using hyperspectral technology.The specific contents are as follows:(1)Establishment of prediction model of soluble solid content in huangguan pear and dangshansu pear.S-G convolution smoothing method,multivariate scattering correction method(MSC),standard normal variable transformation method(SNV)and detrending method(DT)were used to preprocess the original spectral curves of huangguan pear and dangshansu pear.The linear PLSR and SVR as well as the non-linear RBF-PLSR and RBF-SVR algorithms are selected to establish the prediction model for the preprocessed spectral data.The results show that after pretreatment with SNV method,the prediction model of soluble solids content of huangguan pear fruit established by SVR algorithm is the best.The best model modeling set R2=0.99,RMSE=0.06,prediction set R2=0.94,RMSE=0.89.After pretreatment with SNV,the prediction model of soluble solids content of dangshansu pear was established by SVR algorithm.The modeling set R2=0.99,RMSE=0.08,prediction set R2=0.90,RMSE=5.67.(2)Establishment of prediction models for fruit moisture content of huangguan pear and dangshansu pear.S-G convolution smoothing method,multivariate scattering correction method(MSC),standard normal variable transformation method(SNV)and detrending method(DT)were used to preprocess the original spectral curves of huangguan pear and dangshansu pear.The linear PLSR and SVR as well as the non-linear RBF-PLSR and RBF-SVR algorithms are selected to establish the prediction model for the preprocessed spectral data.The results showed that the best prediction model for fruit moisture of huangguan pear was established by SG+RBF-PLSR.The best prediction model modeling set R~2=0.98,RMSE=0.06,prediction set R~2=0.89,RMSE=0.35.After pretreatment with SNV,the RBF-PLSR algorithm was used to establish the Dangshansu pear fruit moisture content prediction model with the best effect.The best prediction model modeling set R~2=0.94,RMSE=0.06,prediction set R~2=0.88,RMSE=0.192.(3)Establishment of the prediction model of fruit hardness of huangguan pear and dangshansu pear.S-G convolution smoothing method,multivariate scattering correction method(MSC),standard normal variable transformation method(SNV)and detrending method(DT)were used to preprocess the original spectral curves of huangguan pear and dangshnasu pear.The linear PLSR and SVR as well as the non-linear RBF-PLSR and RBF-SVR algorithms are selected to establish the prediction model for the preprocessed spectral data.The results showed that the best prediction model of huangguan pear fruit hardness prediction model was established by SG+RBF-PLSR.The modeling set R~2=0.99,RMSE=0.06,prediction set R~2=0.89,RMSE=0.49.The best prediction model of dangshansu pear fruit hardness was established by SNV+RBF-PLSR,the modeling set R~2=0.99,RMSE=0.07,prediction set R~2=0.90,RMSE=1.04.Based on the above research results,the best model pretreatment method and modeling algorithm of the prediction model for the detection of soluble solids content,moisture content and hardness of huangguan pear and dangshansu pear based on hyperspectral technology can be obtained.Provide a basis for non-destructive testing of pear quality in the later period.