Based on near-infrared spectroscopy and chemometrics methods， a non-destructive classification method for different brands and counterfeit soybean milk powder was established in this paper. A total of 132 spectra of counterfeit， domestic and foreign brands were collected. Continuous wavelet transform method was used for the elimination of background and baseline interference. Characteristic wavelengths selection was achieved based on the standard deviation and relative standard deviation. Then principal component analysis method was applied to analyze the spectra of different brands and counterfeit samples. The results show that the baseline interference was effectively eliminated by the continuous wavelet transform method， and discrimination ability was greatly improved by the wavelength screening method. The classification accuracy among counterfeit， domestic and foreign brands is 100%， and the classification accuracy of different brands samples is 93.94%. The results show the classification of counterfeit and different brands of soybean milk powder samples can be achieved by the proposed method， which provides a new idea for the rapid and non-destructive analysis of soybean milk powder samples.