Quantitative analysis of residues of chlorpyrifos in corn oil based on Fourier transform near-infrared spectroscopy and deep transfer learning

Abstract

This study introduces a novel deep transfer learning (DTL) approach based on convolutional neural networks (CNN) and Fourier transform near-infrared (FT-NIR) spectroscopy to enhance the accuracy of predicting the residual levels of chlorpyrifos in corn oil. The method proposed involves creating a 1D-CNN model using existing data and utilizing DTL to enhance the performance of a new model by transferring the parameters learned from the trained model. The research findings demonstrate that compared to the CNN model, the proposed TL method achieves superior predictive accuracy while utilizing a smaller amount of FT-NIR spectral data, thus reducing the reliance on labeled FT-NIR spectral data for model training. Specifically, the coefficient of determination (RP2) reaches 0.9754, and the relative percent deviation (RPD) is 6.4575. This study confirms that the developed DTL method based on CNN and FT-NIR provides an efficient and precise approach to detecting food safety issues. In addition, this method is not only applicable for pesticide residue detection in edible oils based on near-infrared spectroscopy but can also be used for the chemometric analysis of FT-NIR spectral data in other fields.