Quantitive Detection Of Organophosphorus Pesticide Residue In Leaf Vegetables Based On Near Infrared Spectroscopy

With the continuous development of Chinese national economy, food safety has become a crucial issue. In food quality inspection report, pesticide residues are detected in some food and are seriously beyond the standard in vegetables, which does great harm to people’s health. Therefore, research on the technology of pesticide residues detection with less time-consuming, low-cost and environmentally low-carbon is developed both at and abroad.Among all the pesticide, organphosphorus pesticide is the most widely used in agriculture for its high efficiency in recent years. Enhancement on organphosphorus pesticide detection and the research of Environmental Toxicology plays a significant guiding role on protecting people’s health and reducing the incidents of food poisoning caused by pesticide residues.Gas Chromatograph with Flame photometric detector or NPD is currently widely used to detect organphosphorus pesticide based on national standard, but its pretreatment takes a long time with various steps and many complicated interfering substances may exist in the samples. In this way, it is quite difficult to fully exact multiple pesticide residues and get rid of impurity at the same time during pretreatment.This paper first determines the true value of pesticide residues in vegetable samples by GC and then do a research on the method of detecting pesticide residues in vegetable by Near Infrared Spectra in order to establish the model for the predicted and true value of pesticide residues spectrum, commonly used dichlorvos and methamidophos as research object, and Chinese cabbage as research vegetable. In this paper, a sample of350-1800nm near infrared spectra is collected by Visible-near infrared spectrometer, sampling by every1nm and10times of scanning. Meanwhile, the best pretreatment methods, first derivative and FD combined with Standard normal variable transformation, are obtained through comparing the pretreatment on samples by several common spectrums. Near infrared spectra model for the sample is established by Partial least squares method, correlation coefficient between predicted and true value of methamidophos calibrating samples is0.8668, root mean square error is4.4316; correlation coefficient between predicted and true value of forecast samples is0.8476, root mean square error is4.3740; correlation coefficient between the predicted values and the true value of dichlorvos calibrating samples is0.8501, root mean square error is0.8006, correlation coefficient of forecast set r is0.8607, root mean square error after interactive validation is0.2358.This experiment also establishes the correction model of the sample by Support Vector Machine, interval Partial least squares method and Successive project algorithm in order to further improve its accuracy. Correlation coefficient between predicted and true value of methamidophos samples is measured as0.81996by SPA, RMSEP is0.00018; correlation coefficient between predicted and true value of dichlorvos calibrating sample is measured as0.8281by SVM; correlation coefficient between predicted and true value of dichlorvos calibrating sample is measured as0.8256by iPLS, RMSEP is5.0488, correlation coefficient of forecast set r is0.8737, root mean square error after interactive validation is5.3579; correlation coefficient between predicted and true value of dichlorvos r is measured as0.84153by SPA, RMSEP=1.94510.This paper indicates Near Infrared Spectra technology is less time-consuming without damage to the sample and permits multiple analyses concurrently. This technology will be widely used to detect pesticide residues in vegetables in future and this paper offers helpful directions and ideas for the development of technology on pesticide residues detection.