Abstract: This study aimed to achieve a rapid determination for contents of polyphenols and flavonoids in pomegranate peels.And 102 samples collected from Lintong district of Shaanxi Province were determined by near infrared reflectance spectroscopy. To acquire best mathematical models, methods with sample division, pretreatment and band selection were compared. The results showed that the best model of polyphenols was built with sample division using SPXY, spectral pretreatment using multiple scattering correction + Savitzky-Golay derivatives ( 2, 4, 3) , removing outliers by Monte Carlo cross validation ( MCCV) , and effective wavenumbers chosen by competitive adaptive reweighted sampling ( CARS) . The coefficients of determination of calibration ( R²_C) and prediction ( R²_P) for the best model were 0.9699 and 0.9341.The best model of flavonoids was built with sample division using SPXY, spectral pretreatment using Savitzky-Golay Derivatives ( 2, 4, 3) , removing outliers by MCCV, and effective wavenumbers chosen by CARS. The coefficients of determination of calibration ( R²_C) and prediction ( R²_P) for the best model were 0.9775 and 0.9540.Above results indicate that polyphenols and flavonoids contents in pomegranate peels could be determined quickly by near infrared reflectance spectroscopy.