Enhanced detection of aflatoxin B1 in single peanut kernels using laser-induced fluorescence and a weighted algorithm

Peanuts, a globally significant crop, are prone to aflatoxin B₁ (AFB₁) contamination, posing a significant threat to food safety. This study employed laser-induced fluorescence spectroscopy (LIFS) to detect AFB₁ in single peanuts. Natural contamination conditions were simulated to obtain peanuts with different AFB₁ levels, and surface fluorescence signals were collected using single-probe and three-probe methods. Toxin content was quantified through wet chemistry, and machine learning was applied for classification. The results showed that increasing the number of probes significantly improved detection accuracy and reduced the false negative rate (FNR). A weighted algorithm was proposed to enhance spectral analysis, which can amplify the differences between contaminated and uncontaminated samples. A linear SVM based on the three-probe weighted fluorescence spectral data achieved best discriminant ability (accuracy = 100%). Additionally, the Random Forest (RF) algorithm identified six key wavelengths, enabling an SVM classifier to predict contamination with 94.12% accuracy and a 0% FNR. This high-sensitivity, high-accuracy method provides a reliable technical solution for rapid, nondestructive AFB₁ detection in peanuts, offering promise for critical applications in food safety monitoring.