Nondestructive Detection Methods Of Some Peanut Quality Parameters Based On Hyperspectral Imaging Technology

Peanut is one of the most important food crops in China,which has very high economic and nutritional value and plays an important role in the whole food market.The quality of peanut is directly determined by its moisture content,fat content and mildew.However,the traditional detection methods are mostly chemical methods,which need to destroy peanut samples.The detection processes are time-consuming and complicated,and even largely influenced by the experience of the operators.So,it is very important to realize the accurate and rapid nondestructive detection of peanut quality.In this study,peanut was taken as the research object.The accurate and rapid nondestructive methods for detection of moldy peanuts,moisture content and fat content were proposed by using hyperspectral imaging technology combined with chemometric analysis methods.The specific research contents and results were as follows:Firstly,the hyperspectral imaging technology at 1000?2500 nm was used to classify and detect moldy peanuts.The hyperspectral images of all peanut samples were collected and the average spectral data of region of interest(ROI)of each moldy peanut and healthy peanut were extracted.Successive projections algorithm(SPA)was adopted to select effective wavelengths and three classification models including partial least squares discrimination analysis(PLS-DA),support vector machine(SVM),and linear discriminant analysis(LDA)were established based on these effective wavelengths.By comparing,among the three models,SPA-LDA classification model obtained the highest detection accuracy of 100%.Finally,the SPA-LDA classification model was applied to discriminate and classify all the pixels in the hyperspectral image of peanut samples.The moldy pixels were displayed in red and the healthy ones were displayed in green,thus realizing the visual detection of moldy peanuts.Secondly,the hyperspectral imaging technology at the wavelength of 400?1000 nm was utilized to detect peanut moisture content quantitatively.The hyperspectral images of all peanut samples were collected and the average spectral data of each peanut's region of interest(ROI)were extracted.Then,the moisture content of the corresponding peanut samples was measured according to the direct drying method stipulated in the national standard GB 5009.3-2016.Regression coefficient(RC)and successive projections algorithm(SPA)were adopted to select effective wavelengths.Afterwards,based on the full and effective wavelengths,principal component regression(PCR),partial least squares regression(PLSR)and support vector regression(SVR)were established respectively.By comprehensive comparing,SPA-SVR detection model showed the best performance with determination coefficient(R_p~2)of 0.9363,root mean square errors(RMSEP)of 0.7021%and residual prediction deviation(RPD)of 3.988 in the prediction set.Finally,the hyperspectral imaging technology at wavelength of 1000?2500 nm was used to detect peanut fat content quantitatively.The hyperspectral images of all peanut samples were collected and the mean spectral data of each peanut's ROI were extracted.Then,the Soxhlet extraction method specified in the national standard GB5009.6-2016 was used as the standard chemical analysis method to detect the fat content of peanut samples.The fat content of the samples was detected to obtain the actual fat content of each peanut sample.Regression coefficient(RC)and successive projections algorithm(SPA)were adopted to select effective wavelengths.Then,based on the full and effective wavelengths,partial least squares regression(PLSR),support vector regression(SVR)and multiple linear regression(MLR)were established respectively.By comprehensive comparing,SPA-MLR detection model showed the best performance with determination coefficient(R_p~2)of 0.9315,root mean square errors(RMSEP)of 0.4895%and residual prediction deviation(RPD)of 4.0449 in the prediction set.The above researches provided a theoretical basis for the establishment of a set of methods to realize the nondestructive detection of peanut multiple quality parameters systematically.