The Detection Methods Analysis Of Pesticides Residue On The Surface Of Fruits By Raman Spectroscopy

With the progress of national economy and the improvement of people’s life, the qualityand safety issues of food are becoming the focus of attention by people. Routine analyticalmethods used for the determination of pesticides residue on the surface of fruits aredestructive, complex, time-consuming, high cost and not environmentally friendly. Hence, itwill be very worthful for pratical applications to explore a rapid, simple and nondestructivedetection techniques. Selecting Gannan navel orange as the research object and commonpesticides-chloryrifos, dimethoate and phosmet as analysis indicators of detection, the rapidand quantitative detection of pesticides residue on the surface of fruits was studied usingRaman spectroscopy. And the main results of the thesis were involved:1. Confocal micro-Raman spectra of the chlorpyrifos solutions samples were collectedon different substrates （golden, silver, aluminum and copper）. Different spectral pretreatmentmethods and chemometric resolution methods （Principal Component Regression, PartialLinear Squares, Least Squares Support Vector Machine） were applied to optimize theaccuracy of quantitative models. The results showed that the calibration model built byLS-SVM combined with first derivative preprocessing method was the best one for thespectrum collected on the silver substrate. The correlation coefficient of prediction set（RP）and root mean square error of prediction （RMSEP） were0.996and0.62%.2. The feasibility of rapid quantitative determination for pesticides residue on the surfaceof fruit was discussed using Raman spectroscopy. Different spectral pretreatment methods andchemometric resolution methods （PCR, PLS, LS-SVM） were applied to optimize the accuracyof quantitative models. The results showed that the model built by PLS combined with SNVwas ideal with0.995of RP,0.77%of RMSEP.3. The SERS spectra of dimethoate （0.5μg/ml～10μg/ml） and phosmet （0.1μg/ml～10μg/ml） were collected using confocal micro-Raman spectroscopy and the quantitativemathematical models of the two kinds of pesticide solutions were established. Differentspectral pretreatment methods were compared and ideal the modeling wavelengths wereanalyzed for the models. The research results showed that the model built by PLS combinedwith first derivative and standard normal variable transformation （SNV） is the best one in the ideal wavelengths(1519.5～1850.3cm^-1、1189.8～1355.6cm^-1、694.2～1024.9cm^-1and199.5～530.2cm^-1) for dimethoate solutions. The correlation coefficient of calibration （RC）and prediction （RP） were0.980and0.951. The root mean square error of calibration （RMSEC）and prediction（RMSEP） were0.652μg/ml and0.788μg/ml respectively. The model built byPLS combined with first derivative is the best one in the ideal wavelengths(1429.9～1850.3cm^-1、1150.2～1291cm^-1and450.2～1010.4cm^-1) for phosmet solutions with0.982of RC,0.647μg/ml of RMSEC,0.957of RP,0.655μg/ml of RMSEP.The detection of pesticides residue was analyzed by Raman spectroscopy and SERS.And the results indicate that it is feasible to make quantitative detection of pesticides residueon the surface of fruits with Raman spectroscopy combined with chemometrics.