In this work, a quick and quantitative detection method for three common flavors, i.e., coumarin, vanillin, and ethyl maltol, in flavor adulterated Pu'er tea is established. The Fourier transform near-infrared spectroscopy combined with the partial least squares method is used to quantitatively analyze the flavor adulterated Pu'er tea. The quantitative analysis models for three adulterated flavor components are established, and the predictive capabilities of the quantitative analysis models built with different pre-processing methods and without spectral pre-processing are compared. The results show that the predicted root mean square errors of the three flavors of coumarin, vanillin, and ethyl maltol by combining different spectral pre-processing methods are 0.1461, 0.1678, and 0.1800, respectively, and the prediction determination coefficients are 0.7989, 0.7350, and 0.6938, respectively. The detection limit of three flavors is 0.2 mg/g. The near-infrared spectroscopy combined with the partial least squares quantitative analysis can achieve rapid detection and analysis of the three flavors in adulterated Pu'er tea.