Detection Of Multiple Pesticide Residues In Lettuce Leaves Based On Hyperspectral Image Technology

Because of the serious threat to people’s health,pesticide residue has become a problem of food safety that can not be ignored.Therefore,it is very necessary to realize the nondestructive and accurate detection of pesticide residues.There are a wide variety of traditional methods for pesticide residue detection,but the defects of destructive to samples and the complex pretreatment make them unfavorable for field detection.The sample information obtained by NIRS,RS and FS is not comprehensive enough for their point source sampling,which causes that the detection result of pesticide residues is not ideal.And researches are mainly aimed at the analysis of single pesticide residue in the sample,but the effect of single pesticide prevention and control in real life is often not significant.The phenomenon of mixed use of pesticides is becoming more and more common,which makes it difficult to meet the detection of multiple pesticide residues by means of nondestructive detection for single pesticide residue.Therefore,it is very important to study a nondestructive and accurate method to detect multiple pesticide residues.In this paper,lettuce leaves were used as the research object,and hyperspectral image technology was used to detect multiple pesticide residues.The main contents and conclusions are as follows:(1)The characteristics of spectral curves of lettuce leaves were analyzed by combining the existing researches,which proved that hyperspectral image technology is feasible for identifying multiple pesticide residues and quantitatively detecting mixed pesticide residues in lettuce leaves.(2)Hyperspectral image technology was used to identify multiple pesticide residues in lettuce leaves.Firstly,hyperspectral image system was used to acquire hyperspectral images of leaf samples for extracting spectral data.Secondly,descriptive statistics and SNV algorithm were successively used to pretreat the original spectra,and SPA algorithm was used to select 15 characteristic wavelengths.Thirdly,SVM models were established based on full spectral data and characteristic spectral data respectively,which indicated that the latter SVM model has better effect.Then three kinds of intelligent algorithms of ABC,CS and GSA were used to optimise SVM parameters,and the SVM model optimised by GSA(GSA-SVM)was optimal.Finally,the identification effect of GSA-SVM models under different particle numbers(N)was studied to find that GSA-SVM model when N was 40 achieved the best effect,with accuracy of 100%for training set and 96.08%for test set.Therefore,it is feasible to accurately identify multiple pesticide residues in lettuce leaves by hyperspectral image technology.(3)Hyperspectral image technology was used to quantitatively detect mixed pesticide residues in lettuce leaves.After acquiring hyperspectral images of leaf samples,the mixed pesticide residues were measured by GC-MS,and original spectral data was extracted by ENVI.The SNV algorithm was used to pretreat the original spectra,then characteristic wavelengths were selected by CARS and RF-RFE.And the LSSVR models of each pesticide residue were established based on full spectral data and characteristic spectral data.It was found that CARS-LSSVR model could detect fenvalerate residue better,and RF-RFE-LSSVR model could detect dimethoate residue better.Finally,SPA algorithm was used for secondary wavelength selection to reconstruct LSSVR models for two pesticide residues.The results showed that CARS-SPA-LSSVR model of fenvalerate residue reached R_P~2 of 0.8890 and RMSEP of 0.0182,and RF-RFE-SPA-LSSVR model of dimethoate residue reached R_P~2 of 0.9386 and RMSEP of 0.0077,which indicated that quantitatively and accurately detecting mixed pesticide residues in lettuce leaves by hyperspectral image technology is feasible.In this paper,hyperspectral image technology was used not only to identify multiple pesticide residues in lettuce leaves,but also to realize the quantitative detection of mixed pesticide residues.It provides a method basis for the detection of multiple pesticide residues in other vegetables,and also lays a foundation for the development of a nondestructive and accurate detecting instrument for multiple pesticide residues in vegetables.