Evaluation of different models for non-destructive detection of Profenofos residue

Abstract

Pesticides are widely used in agricultural practices to protect crops from pests and diseases. However, their usage can also pose a risk to human health and the environment, especially when their residues remain in food products. Therefore, the development of accurate and efficient detection methods for pesticide residues is crucial for food safety. The aim of this study was to investigate the effectiveness of an artificial intelligence approach utilizing a machine learning algorithm to detect the presence of profenofos residue in samples using hyperspectral imaging-based image processing. Hyperspectral imaging is a non-destructive method that can rapidly acquire spectral information over a wide range of wavelengths, providing a unique spectral signature for each sample. A portable UV-VIS-NIR spectrometer was utilized to measure treated filter papers and the reflectance values of the sample were analyzed at specific wavelengths. Six machine learning algorithms were evaluated. The inputs used in the ANN model were RGB values and intensity of the tested samples. The results showed that ANN was the most accurate algorithm, with 74.4% accuracy in 64 MP camera, 69.9% in 32 MP camera and 66.7% in 13 MP camera. Additionally, the study found that a higher camera resolution, such as 64 MP, produced better processing results due to its ability to capture more detailed images. Overall, this research highlights the potential of machine learning and hyperspectral imaging for the detection of pesticide residues in food products.