Identification of pesticide residues in lettuce leaves based on near infrared transmission spectroscopy

To explore the nondestructive and accurate qualitative detection method of pesticide residues in vegetables, lettuce leaves were taken as carrier and pesticides (fenvalerate and chlorpyrifos) were taken as research object. The near infrared spectral data of lettuce leaves was collected by using near infrared spectrometer NIRez (950-1,650 nm), then it was preprocessed by SNV algorithm and SG-SNV combination, respectively. The successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS), iteratively retaining informative variables (IRIV), and combination of two of them were used to select characteristic wavelengths from the preprocessed spectral data, respectively. In this study, support vector machine (SVM) was used to classify pesticide residues in lettuce leaves. Because the classification performance of SVM was influenced by parameters c (penalty coefficient) and g (kernel function parameter), so the parameters c and g of SVM were optimized by gravitational search algorithm (GSA). Finally, identification models were built on the full wavelengths and the selected characteristic wavelengths, respectively. Final result showed that CARS-IRIV-GSA-SVM model had the best performance, and the classification accuracy of training set and predication set reached 100 and 98.33%. Therefore, near infrared transmission spectroscopy can be applied to the qualitative detection of pesticide residues in lettuce leaves.