Research On The Inspection Of Common Herbicides Residue On Mint Leaves Based On Various Spectrum Methods

Herbicides have a large share in the pesticide market,with high market sales,wide application range and various types.In the public security work,cases of excessive application of herbicides to cash crops to achieve infringement retaliation or disputes caused by wrong use occur from time to time.Spectrum method is the main method for the detection of pesticide residues,among which spectral method and LC-MS are the most widely used.In the first part of the experiment,the species identification model of herbicide residues on mint leaves based on hyperspectral imaging and chemometrics was established,establish the classification model,and evaluate the classification results: 1.The image information of mint leaves is compared and analyzed to form a pseudo color image,and the first three principal components of the image are extracted by principal component analysis to form an RGB image.2.Analyze the curve information.The classification and recognition performance of different spectral preprocessing methods,dimensionality reduction methods,classification models and parameter optimization algorithms for herbicide species were compared.The experimental results show that: 1.The pseudo color images and RGB images of mint leaves treated with different herbicides are close,which is difficult to distinguish,but the diseased spots are more prominent and easier to observe after PCA analysis.2.The recognition accuracy of SVM can reach 96.53%,which is significantly better than ELM mechanical learning machine.3.SG and MSC have little effect on the accuracy of model discrimination.SPA is better than PCA in preserving the original spectral information.PSO is excellent and stable in the process of parameter optimization,and the calculation time is relatively short.In the second part of the experiment,a qualitative identification method of 45 sulfonylurea and amide herbicides based on HPLC-MS/MS was established: 1.Optimize the conditions to realize the simultaneous detection of multiple herbicides and determine the detection limit.2.Use this method to screen and test the types of herbicides left on cabbage,radish and flower leaves in the actual case.The experimental results show that: 1.The elution and response of 0.1‰ammonia acetonitrile,0.002 mol/L ammonium formate-0.1 ‰ formic acid acetonitrile and 0.1‰formic acid acetonitrile were compared.0.1‰ formic acid acetonitrile was selected as the mobile phase of this experiment.The mass spectrometry conditions were optimized to identify the types of herbicides,and the detection limit was 0.05 ng/m L ～ 2.00 ng/m L.2.The screening and inspection of herbicide residues on the leaves of three kinds of plants in the actual case were successfully realized.The type of residual herbicide is rimsulfuron.At the same time,the hyperspectral imaging combined with HPLC-MS/MS was used to evaluate the identification results of herbicide residue concentration,and the hyperspectral information of peppermint leaves applied with low,medium and high types of metolachlor was collected.On the basis of the first and second experiments,the third experiment was completed:1.The SG-SVM model is established according to the hyperspectral curve,and the effects of GS,PSO and GWO optimization algorithms on the accuracy of model recognition are compared.2.Establish a quantitative identification method of alachlor based on HPLC-MS/MS to detect the actual concentration of experimental samples.3.Compare the actual concentration and classification results of misclassification samples in the classification and recognition model,and evaluate model capability.The experimental results show that: 1.After extracting the principal components of hyperspectral image information,the difference between different concentrations is obvious,and the disease spots are clear and identifiable.2.The highest classification accuracy of SG-GS-SVM model training set and test set is 98.67%.3.The quantitative determination method of metolachlor based on HPLC-MS/MS showed that the average content of metolachlor in mint leaves with low,medium and high concentrations was152.04ng/mg,353.24ng/mg and 1518.68 g/mg,respectively.4.Combined with the concentration detection results of HPLC-MS/MS,the recognition effect of classification and recognition model should pay attention to the classification results of overall samples.SG-GSSVM has the strong generalization ability for special samples.In this paper,hyperspectral imaging and HPLC-MS/MS are used to realize the classification and identification of herbicide residues on mint leaves,which has a good classification ability,improves the existing technical methods and enriches the method system of pesticide residue detection.