Near-infrared spectroscopy (NIRS) quantitative detection models of sesame oil adulterated respectively with soybean oil, corn oil and peanut oil were established by interval partial least squares (iPLS). Mixed samples of sesame and adulterated oils with different proportions were scanned by using a FT-NIR spectrometer and the near infrared transmission spectra were collected in 4000 ~12000cm-1 region which divided into calibration and prediction set. The spectra in 4420~12000cm-1 region was pretreated with different ways and the best was Savitzky—Golay moving. The characteristic wavelength bands were selected by iPLS and the quantitative models were established by PLS finally. The results showed that predicted correlation coefficients of the models built with PLS for three kinds of vegetable oil reached to 0.998，0.999，0.999 respectively, while the root mean square error of prediction were 0.24%，0.24%，0.19%, which have high accuracy. The experiments proved that NIRS technique used for rapid detection of adulterated oil in sesame oil possesses certain feasibility.