In order to evaluate the ability of near infrared spectroscopy (NIRS) to detect moldy core in apple fruits, the discriminant efficiency of Fisher function based on several spectra preprocessing methods and principal component analysis (PCA) were investigated. Results showed that the discriminant efficiency of Fisher function established by 20 principal components (PCs) after vector normalization preprocess was the highest, discriminant efficiency for training sample set and test set were 89.9% and 87.8%, respectively.