No abstract

To explore the best method for non‐destructively and accurately quantitative detection of mixed pesticide residue in vegetables, the mixed pesticide (fenvalerate and dimethoate) of lettuce leaves was used as the research object detected by hyperspectral technique. The hyperspectral data was preprocessed using standard normalized variable. Then two different kinds of characteristic wavelengths were selected using competitive adaptive reweighed sampling (CARS) and random forest‐recursive feature elimination (RF‐RFE), respectively. The least squares support vector regression (LSSVR) model results of predicting fenvalerate and dimethoate showed that the CARS‐screened wavelengths had the best modeling results for fenvalerate with RP2 of 0.8203 and RMSEP of 0.0222, and the RF‐RFE‐screened wavelengths had the best results for dimethoate with RP2 of 0.8712 and RMSEP of 0.0186. To simplify the calibration model, successive projections algorithm (SPA) was used for the second selection of characteristic wavelengths. Finally, the CARS‐SPA‐LSSVR model for predicting fenvalerate achieved the accuracy with RP2 of 0.8890 and RMSEP of 0.0182, RF‐RFE‐SPA‐LSSVR model for predicting dimethoate with RP2 of 0.9386 and RMSEP of 0.0077. Thus, hyperspectral technique can be applied to quantitatively detect the mixed pesticide residue of lettuce leaves. Practical applications Lettuce leaves can be used for raw food, and lettuce leaves with mixed pesticide residues is more likely to bring harm to people's health. Traditional methods of detecting pesticide residues are time‐consuming and destructive to samples, so they cannot meet the requirement of modern agriculture. Whereas, hyperspectral imaging technology is a new rapidly growing method which integrates spectroscopic and imaging techniques in one system for providing both spectral and spatial information simultaneously, of which the spectral information can detect the physical structure and chemical composition of unknown samples. Therefore, the hyperspectral technique is a fast and nondestructive detecting method, and this study showed it is feasible for quantitative detection of the mixed pesticide residue of lettuce leaves.

A rapid and nondestructive method based on the VIS-NIR hyperspectral imaging technique in the wavelength range of 390-1050 nm was investigated for discriminating the varieties of black beans.Totally 300 samples of three varieties were scanned by the VIS-NIR hyperspectral imaging system, and hyperspectral data were analyzed by spectral and image processing technique respectively. Successive projections algorithm (SPA) was used to obtain 13 characteristic wavelengths (504, 507, 512, 516, 522, 529, 692, 733, 766, 815, 933, 998 and 1000nm) for spectral analysis. After the processing of SPA, optimal image selection was carried out by Principal component analysis (PCA) based on the characteristic wavelengths. The first principal component (PC1) image was used for the image analysis, whose contribution rate was over 98.34%.Gray-level co-occurrence matrix (GLCM) analysis from PC1 image was applied to extract image Downloaded by [University of Otago] at 01:47 14 July 2015 A c c e p t e d M a n u s c r i p t 2 features including 16 textural features and 6 morphological features. In this study, Partial Least Squares-Discriminate Analysis (PLS-DA), Support vector machine (SVM) and K-nearest neighbors (KNN) were used for model establishments respectively based on spectral feature, image feature, and the combination of spectral and image features. The results show that the best correct discrimination rate of 98.33% was achieved by applying combined spectral and image features. The study demonstrated that VIS-NIR hyperspectral imaging technique was potential for rapid classification of black beans, and the performance of the classification model can be improved by the feature combination.

To facilitate more quickly and effectively detect the types of pesticide residues on the surface of lettuce, a method involving the chemical molecular structure coupled with wavelet transform (CMS‐WT) was proposed to extract the characteristic wavelength. Five different kinds of pesticide residues were sprayed on the surface of lettuce, respectively, dimethoate, acephate, phoxim, dichlorvos, avermectin (the ratio of pesticides and water were 1:1000). In addition, the near infrared hyperspectral image information of 200 samples in five different kinds of pesticides residue in lettuce were achieved by the NIR hyperspectral imaging system (870–1780 nm). The region of interest (ROI) in hyperspectral image of samples was selected to get the near infrared spectral data by the software of ENVI. Furthermore, CMS‐WT was used to extract the most influential wavelengths. Four characteristic intervals were extracted by comparing the different of pesticides in chemical molecular structures, respectively, 875—1050 nm, 1050—1250 nm, 1350—1550 nm, 1650—1780 nm. Further, the best combination of eight features were selected according to the reorder of the size of the singular value by wavelet transform algorithm using db6 as wavelet basis function, respectively, 919.18, 944.25, 972.25, 1194.20, 1363.81, 1426.69, 1673.29, 1680.74 nm. Finally, SVM model was established according to the extracted characteristic spectral data. The results showed that the calibration and prediction accuracy of SVM model established by the best combination of eight features were all achieved 100%. It confirms that the CMS‐WT feature extraction algorithm is feasible and effective for building models of different pesticide residues in lettuce. Practical applications Well understanding the effect of pesticide residues to biological structure is very important for revelation of novel biological function and mechanism of action of the protein. To facilitate more quickly and effectively detect the types of pesticide residues on the surface of lettuce, a method involving the chemical molecular structure coupled with wavelet transform (CMS‐WT) was proposed to extract the characteristic wavelength in this article. It confirms that the CMS‐WT feature extraction algorithm is feasible and effective for building models of different pesticide residues in lettuce.

No abstract