In order to find out a fast measure method of adulterated milk based on near infrared spectroscopy, milk adulterated with plant butter, vegetable protein and starch was collected respectively. Using Fourier transform near infrared spectroscopy to scan the samples, the spectrum data were obtained. The samples were scanned in the spectral region between 4 000 and 12 000 cm-1 by FT-NIR spectrometer with an optic fiber of 2 mm path-length and an InGaAs detector. Then all data were analyzed by principal component analysis combined with Fisher line discriminant analysis (FLDA) and partial least squares (PLS). Results show that the accumulative reliabilities of the first six components were more than 99%, so the first six components were applied as FLDA inputs and the values of the type of milk were applied as the outputs. An adulterated milk qualitative discriminant model based on Fisher line discriminant analysis was developed finally. The result indicated that the accuracy of detection of calibration samples is 97.78%. The unknown test samples were tested by this model and the correct identification rate is 94.44%. Partial least square models for detecting the content of material added to raw milk were set up with good veracity. The predictive correlation coefficient (R2) of calibration sets of milk adulterated with plant butter, vegetable protein and starch are 99.08%, 99.96% and 99.39%, respectively, while the root mean square errors of cross validation (RMSECV) of the three calibration sets are 0.304%, 0.013 5% and 0.060%, respectively. The R2 of validation sets of the three kinds of adulterated milk are 98.50%, 99.94% and 98.50%, respectively, while the root mean square errors of prediction (RMSEP) of the three validation sets are 0.323%, 0.028 8% and 0.068%, respectively. All of these suggested that near infrared spectroscopy has good potential for rapid qualitative and quantitative detection of milk adulterated with botanical filling material.