This paper studied the feasibility to detect talcum powder illegally added in tea based on Fourier transform infrared (FTIR) transmission spectroscopy with chemometric methods. In this study, 210 tea samples with 12 dose concentrations of talcum powder were prepared for FTIR spectra acquirement. Firstly, Savitzky-Golay (SG) smoothing, normalize and standard normal variate (SNV) were used to preprocess the raw spectra. It was shown that SNV preprocessing had the best performance. After that, a hybrid method of backward interval partial least squares (biPLS) regression and successive projections algorithm (SPA) was used to select 5 characteristic wavenumbers, which only accounted for 0.18% of the whole wavenumbers. Then, PLS regression and least square support vector machine (LS-SVM) were utilized to build linear and nonlinear models based on these 5 characteristic wavenumbers, respectively. Finally, the optimal model was achieved with LS-SVM with highRP=0.921 and low RMSEP=0.131. It concluded that talcum powder in green tea could be detected based on FTIR spectroscopy coupled with chemometrics.