Based On Hyperspectral Imaging Technology Non-destructive Testing Method For Pesticide Residues On Fruit Surface

Fruit is an indispensable part of people’s daily diet and a high-risk area for food safety issues.This paper studies the problem of pesticide residues on fruits in food safety.At present,the detection methods of common pesticide residues in fruits and vegetables mainly include liquid chromatography,gas chromatography,enzyme inhibition,supercritical fluid extraction(SFE),and biosensor methods.They are not only complicated to operate,time-consuming,economical,but also destroy the sample structure,which is a lossy detection method.Therefore,there is an urgent need for a non-destructive,non-contact,simple,rapid and accurate method for detecting pesticide residues on fruits and vegetables.Based on hyperspectral imaging technology,this paper studies the non-destructive detection methods of pesticide residues on the surface of persimmon and different pesticide residues on the surface of navel orange.The main research contents of the thesis are as follows:(1)Three kinds of pesticides,such as pyrimethanil,sputum and chlorpyrifos,were sprayed on the surface of persimmon and different concentrations of scorpion pesticide solution were sprayed on the surface of navel orange.A group of samples sprayed with distilled water were established as the control group.Divided into four groups and collected hyperspectral images,determined the region of interest(ROI)and extracted spectral information,and preprocessed the sample raw spectra using standard normal variable(SNV).(2)To study the detection methods of different pesticide residues on the surface of persimmon based on hyperspectral technology.The hyperspectral data of four sets of samples were collected,and Principal Component Analysis(PCA)and Continuous Projection(SPA)were used to reduce the dimension under the full band and extract the spectral feature bands.The characteristic wavelengths(522nm,670 nm,700nm,800 nm,940nm)extracted by PCA and the characteristic wavelengths(522nm,644 nm,685nm,720 nm,790nm,930nm)extracted by Successive projections algorithm(SPA)are respectively supported.Vector machine(SVM),BP neural network and extreme learning machine(ELM)pesticide residue species detection model.The experimental results show that the SPA is better than the PCA.The feature extraction based on continuous projection method is the best.The recognition accuracy is 90%.(3)To study the detection methods of different residual concentrations of pesticides on navel orange surface based on hyperspectral technology.The hyperspectral data of four sets of samples were collected,and PCA and SPA were used to reduce the dimension under the full band and extract the spectral feature bands.Characteristic wavelengths(500nm,580 nm,680,850 nm,930nm,980nm)extracted by PCA and spectrally selected wavelengths extracted by SPA(461nm,549 nm,600nm,680 nm,740nm,830 nm,900nm,960nm)respectively established support vector machine(SVM),BP neural network and extreme learning machine(ELM)pesticide residue concentration detection model,experimental results show that the feature extraction based on continuous projection method combined with support vector machine classification is the best,identification The accuracy rate is 89.58%.