Abstract
To distinguish different adulterated vegetable oils in safflower seed oil, counterfeit vegetable oils were identified and quantified. The near-infrared hyperspectral reflectance curve data of four oil samples were obtained in the range of 387–1035 nm. The spectrum data was preprocessed by five methods and combined with four algorithm models to predict different types of oil samples. The results showed that the median filtering algorithm combined with the Gradient Boosting Decision Tree model can achieve a 100% recognition rate for four kinds of oil samples.