A discrimination method was developed for the rapid and non-destructive identification of fat oxidation in different dairy products by three-dimensional fluorescence spectroscopy.The fluorescence data matrix was decomposed by parallel factor analysis(PARAFAC) and the photosensitive oxidation elements were found from load vectors.After clustering of different samples using score vectors within components，the method of partial least squares discrimination model for different levels of oxidation was established.The yogurt samples with varying degrees of oxidation were collected from different storage environments，in which the lipid oxidation processes of riboflavin and other fluorophores were obtained.The partial least-squares discrimination model of the different storage stages and oxidation levels was established by score vectors.The model specificity and sensitivity of classification were higher than 88。9%，which validated the method in judging the lipid oxidation level of dairy product.