Identification and content prediction of antibiotics in milk based on fluorescence hyperspectral technology

Antibiotic residues in milk affect the quality and safety of dairy products, which are crucial to the health of consumers and the sustainable development of the dairy industry. This study utilizes fluorescence hyperspectral imaging (F-HSI) technology in combination with machine learning algorithms for the rapid, nondestructive detection of quinolone antibiotics (ciprofloxacin, ofloxacin) and chloramphenicol antibiotics in milk. The results demonstrated that the MSC-whole-band-SVM and MSC-RF-SVM combination classification models performed the best, with accuracies of 100% for the calibration set and 97.5610% for the prediction set. The original spectrum-SPA+2D-COS-PLSR algorithm model was the best quantitative model for ciprofloxacin residue samples, with an R2 P of 0.88167 and an RMSEP of 0.030637. The MSC-RF + 2D-COS algorithm model was the best quantitative model for ofloxacin residue samples, with an R2 P of 0.89832 and an RMSEP of 0.030531. The SNV-RF + 2D-COS algorithm model was the best quantitative model for chloramphenicol residue samples, with an R2 P of 0.8755 and an RMSEP of 0.036625. These results indicate that F-HSI combined with machine learning can effectively identify and predict the type and concentration of antibiotic residues in milk, and further evaluation under regulatory standards and practical conditions may enhance understanding of its applicability.