Abstract: In order to quickly detect the quality of frying oil from western-style fast food restaurants, calibration models of acid value and total polar compounds were established using near infrared spectroscopy (NIRS) combined with partial least squares (PLS) to evaluate the frying oil quality. The findings presented that the correction coefficient of calibration model of acid value and total polar compounds were both 0.9974. The root mean square error of cross-validation (RMSEC) were 0.111 and 0.359, respectively. The root mean square error of prediction (RMSEP) were 0.171 and 0.562, respectively. The results of blind samples verification showed that the correlation coefficients between prediction value and true value for acid value and total polar compounds of frying oil were 0.9944 and 0.9761, respectively; the precision test showed that the relative standard deviations (RSD) of acid value and total polar components using the calibration models for the same frying oil sample were 0.934% and 1.278%, respectively, which indicated that the models of quantitative analysis of acid value and total polar compounds of frying oils were good with excellent prediction abilities and reproducibility. Therefore, the rapid detection model based on near infrared spectroscopy could rapidly and accurately detect the quality of frying oil from western-style fast food restaurants.