The Rapid Detection Of Tomatoes Organophosphorus Pesticide Residual Based On Near-Infrared Spectroscopy

Pesticide residues in fruits and vegetables are important risks for food safety, and seriously threat the people’s health. Therefore, rapid detection of pesticide residues in fruits and vegetables is especially important. The existing methods of detecting pesticide residue can not achieve modern needs of quick and green detection, instead, they require complex pre-processing, long time measuring and complicated operation. Near-infrared spectroscopy was used in this paper for rapid detection of tomato organophosphorus pesticide residues in order to provide theoretical basis and technical parameters for development of on site analytical instruments which is rapid in detection.The paper studied the rapid detection of tomatoes organophosphorus pesticide residual basen on near-infrared spectroscopy. Near-infrared spectroscopy was introduced to analyze the tomatoes organophosphorus pesticide residual.10 kinds of spectral pretreatment methods were compared for their effects on model performacne, and both qualitative and quantitative model were obtained for identifying tomatoes organophosphorus pesticide residues; Wavelet was used to eliminate spectral noise before modeling to optimize tomatoes pesticide residual analytical model; The Fourier feature extraction of near-infrared spectroscopy was used to optimize tomatoes pesticide residual analytical model. The main results are as follows:1. Intact tomatoes were used for study, and the BP neural network nonlinear model was used to establish qualitative analysis model. The results show that, by vector normalization, when three principal components and three hidden layers of the network were selected, the best recognition rate of the model was 0.96, with training error of 0.0153, and the crrelation coefficient of forecast value with true value was 0.9711.2. Tomato juice was used for study, and the quantitative analysis model for the tomatoes organophosphorus pesticide residual was set up by PLS based on near-infrared spectroscopy. The result indicated that the model was best when the first derivative preprocessing was selected and the smoothing point was 5 points. The decision coefficient R2 of the near-infrared spectroscopy quantitative model was 0.9836, with a standard error of 0.168% for the interactive calibration model, and the model forecast decision coefficient R2 was 0.9882, with the forecast standard error 0.141% for the interactive calibration model.3. Wavelet was used to eliminate noise of spectrum before modeling. Wavelet could effectively remove the noise information of the spectrum. The decision coefficient R2 of the near-infrared spectroscopy quantitative model was 0.9836, with a standard error of 0.168% for the interactive calibration model, and the model forecast decision coefficient R2 was 0.9882, with the forecast standard error 0.141% for the interactive calibration model. When the decomposition scale was 5, the quantitative analysis model performed best.4. Feature parameters were extracted by Fourier transform and were utilized to build analysis model of tomatoes organophosphorus pesticide residual based on near-infrared spectroscopy. The spectral data of the quantitative analysis were taken for study, and when 35 parameters were choosen, the PLS prediction model, collected from Fourier transform parameters and characteristics of the tomatoes organophosphorus pesticide residual, could be optimal, and the model decision coefficient and RMSECV were 99.18% and 0.118%, respectively; the model forecast decision coefficient R2 and the forecast standard error were 99.06% and 1.118%.