Study on pesticide residues classification of lettuce leaves based on polarization spectroscopy

Abstract

In order to effectively implement the rapid and nondestructive testing of pesticide residues in lettuce leaves, polarization spectral detection technology was used in this article. There were 90 pieces of lettuce leaves from five different groups, as well as a total of 450 samples of lettuce were used to collect polarization spectral information. CARS, IRIV, and SPA were used to obtain optimal wavelengths. BP neural network, KNN, and SVM were used to establish classification models. The results showed that the significance of these parameters from primary factor to secondary factor were incident zenith angle A, detector polarization angle D, detecting zenith angle B, and sample stage azimuth E, respectively. Besides, optimal level of A, B, D, and E were 60°, 45°, 30°, and 270°, respectively. Furthermore, the best classification model different kinds of pesticide residues in lettuce leaves was CARS‐SVM model, with calibration identification rate of 100%, prediction identification rate of 97.78%. It confirms that polarization spectral detection technology is a feasible and effective method for classifying different pesticide residues in lettuce leaves. Practical applications It is of great significance to understand the effects of pesticide residues on the biological structure and to reveal new biological functions and mechanisms of action. In order to effectively implement the rapid and nondestructive testing of pesticide residues in lettuce leaves, polarization spectral detection technology was used in this article. In addition, the order and superior level of polarization spectral factor were obtained through the calculation of the range using orthogonal test. It confirms that the polarization spectra is a feasible and effective method for discriminating different pesticide residues in lettuce leaves.

Nondestructive testing methods for pesticide residue in food commodities: A review

“…Hence, it has to be free from chemical and microbial contamination. Xin et al (2018) proposed the application of polarization spectral detection technology for the detection of pesticide residues in lettuce leaves. The incident zenith angle A, detector polarization angle D, detecting zenith angle B, and sample stage azimuth E were found to have significant relevance from primary to secondary factors.…”

Discrimination of Pesticide Residue Levels on the Hami Melon Surface Using Multiscale Convolution

“…Sun et al [ 12 ] established an optimized support vector machine (SVM) model using near-infrared transmission spectroscopy (950–1650 nm), which could identify two pesticide residue types (fenvalerate and chlorpyrifos) in lettuce leaves, and the prediction accuracy was 98.33%. Zhou et al [ 13 ] used Vis/NIR polarization spectroscopy (300–1000 nm) to identify five pesticide residue types (avermectin, dichlorvos, dimethoate, phoxim, and acephate) in lettuce leaves, achieving a prediction accuracy of 97.78%. Ndung’u et al [ 14 ] used principal component analysis (PCA) to reduce the dimensionality of Vis/NIR spectra (325–1075 nm), and established a machine learning model to identify pesticide residues (mixtures of beta-cyfluthrin and chlorpyrifos, mixtures of metalaxyl and mancozeb) in spinach.…”