A pesticide residue detection model for food based on NIR and SERS.

This paper presents a multivariate calibration model based on Near Infrared Spectroscopy (NIR) and Surface Enhanced Raman Spectroscopy (SERS) techniques, aiming to achieve efficient and accurate detection of pesticide residues in food by integrating the spectral information from both techniques. The study utilizes the Hilbert-Schmidt Independence Criterion-based Variable Space Iterative Optimization algorithm (HSIC-VSIO) for feature variable selection, and combines it with Partial Least Squares Regression (PLSR) to build a spectral fusion quantitative model. Experimental results show that the calibration set Root Mean Square Error (RMSE1) of the NIR and SERS feature-layer fusion model is 0.160, the prediction set RMSE (RMSE2) is 0.185, the prediction set coefficient of determination (R²) is 0.988, and the Relative Percent Deviation (RPD) is 8.290. Compared to single spectral techniques, the NIR and SERS spectral feature-layer fusion method demonstrates significant superiority in detecting pesticide residues in complex matrix samples. The findings further validate the high sensitivity of SERS technology in detecting low concentrations of pesticides and show that the feature-layer fusion method effectively suppresses matrix interference, enhancing the model's generalization ability. This study provides a reliable tool for the rapid and accurate detection of pesticide residues in food and offers new insights into the application of spectral analysis technologies in the food safety field.