AI-powered rapid detection of meat adulteration with terahertz spectroscopy in food and drug control

Introduction Meat adulteration is an increasing problem in food and drug law enforcement, where existing identification methods are inefficient and of limited scope. Methods This study proposes a detection scheme that combines terahertz time-domain spectroscopy (THz-TDS) with artificial intelligence. Time-domain spectra of five common meat species (chicken, beef, duck, mutton and pork) were collected in the 0.2–2.0 THz band, and the absorption coefficient, refractive index and dielectric constant were calculated. Principal component analysis (PCA) was used for feature dimensionality reduction, and support vector machine (SVM) and convolutional neural network (CNN) models were applied to identify meat species and to quantify adulteration content. Results The optical parameters of the five species differed significantly in the terahertz band (P < 0.05, independent-samples t-test). The SVM model achieved 98.3% accuracy for species identification, and the CNN model predicted adulteration content with an error of less than 3%. Discussion The combined THz-TDS and AI approach is non-destructive, rapid and accurate, and can provide a reliable detection method for on-site law enforcement in the public security food and drug environment field.