Spectral Detection Of Pesticide Residues In Lettuce Leaves Based On Wavelet Feature Extraction And Improved Algorithms

Fast identification of pesticide residue in lettuce leaves plays a key role in the test of food safety.At present,the application of spectrum detection technology in the field of pesticide residues had achieved certain results.However,there are high dimensional data and redundant information waiting to be resolved.Wavelet transform is a partial analysis of time(space)frequencies used to obtain the gradual multi-scal zooming of signals through dilation and translation operations,in order to finally achieve the time subdivision at the high-frequency point,and the frequency subdivision at the low-frequency point.Beside,the sensitive wave bands were extracted through the singularity analysis of the high frequency wavelet coefficient curve.In this research,the detection of pesticide residues in lettuce was used as the background,the feature extraction of wavelet transform and the improvement of the algorithm were the main research contents.The main contents and conclusions in this research are as following:(1)Related electron microscopy was used to detect the microstructure of lettuce leaves.It was found that the internal structure of lettuce leaves changed slightly under different concentrations of dimethoate pesticide residue morphology.Moreover,the vegetation reflectance spectra of the near infrared spectrum were mainly related to the arrangement of the leaf cells and the vegetation structure,and the fluorescence spectra were closely related to the chlorophyll content of plants.Study on the micro structure of lettuce leaves sprayed with different concentrations of dimethoate pesticide provided the basis for the study on the mechanism of near infrared spectrum technology and fluorescence spectrum technology to detect the different concentrations of pesticide residues in lettuce.(2)Choosing different wavelet feature extraction parameters combination and its application in pesticide residues in lettuce leaves fluorescence spectral data processing.Besides,5 different algorithms were used to preprocess the raw spectra,respectively.Support vector machine(SVM)classification models were established based on characteristic wavelengths.Furthermore,the wavelet features were selected by wavelet transform(WT)using db4,db6,sym5,sym7 as wavelet basis functions,respectively.The prediction set identification rate of SVM model based on wavelet features obtained the best results compared with the others.Moreover,SG-SNV detrending-WF-SVM model obtained the best performance among all SVM modelswith an identification rate of 98.33% in the calibration set and 93.33% in the prediction set,using sym5 as wavelet basis functions.(3)Combined discrete wavelet transform(DWT)algorithm with the approximate position of frequency doubling center of organic compounds in near infrared spectra,piecewise discrete wavelet transform(PDWT)was proposed in this research,and its application in pesticide residues in lettuce leaves hyperspectral data processing.In order to evaluate the value of the feature extracted by singular value,a parameter of fit degree(FD)was proposed in this research.Combined with the SVM classification accuracy,the feature extracted by PDWT was further evaluated.The classification accuracy of FD,calibration,cross validation and predictive classification accuracy of SVM were,respectively,75%,95%,92.86% and 90.63%,under the N value of 4 with PDWT.(4)WT-MD-MCCV was developed for identifying the optimal wavelengths of the spectral data in this research,and its application in the data processing of the fluorescence spectral data and the hyperspectral data.SG-SNV algorithm was used to preprocess the raw spectra.In addition,PCA,SPA and WT-MD-MCCV were applied to identify the optimal wavelengths.Support vector regression(SVR)was applied to build the prediction models based on preprocessed spectra feature in characteristic wavelengths coupled with different spectral data.WT-MD-MCCV algorithm combined with hyperspectra and chlorophyll fluorescence spectra data performed best among the nine SVR models and the hyperspectra coupled with chlorophyll fluorescence spectra can be used to identify the pesticide residue level in lettuce leaves.The method proposed in this paper can effectively realize the extraction of spectral features and predict the pesticide residues of lettuce leaves in the fast,accurate and nondestructive way.The results of this study can provide certain reference value for the detection of pesticide residues in other crops.