Abstract:To establish a non-destructive identification method for different brands and simple grinding soymilk powder by using near-infrared spectroscopy combined with different optimized pretreatment methods.Firstly，the 132 samples of simple grinding，domestic and foreign brands were collected.The near-infrared original spectra were compared and analyzed.Then，the spectra were processed by different pretreatment methods.Finally，principal component analysis method was used to identify the samples of simple grinding，domestic and foreign brands.The optimal optimized pretreatment method was selected，and a reliable near-infrared identification model for soybean milk powder was established.The results showed that there were obvious background interference and baseline drift in the original spectra，and various pretreatments could effectively eliminate background interference to some extent.The identification of the three kinds of soybean milk powder samples could be achieved by first derivative，continuous wavelet transform，multivariate scatter correction，standard normal variable transformation and the combined pretreatment methods.However，the identification a－mong domestic brands cannot be achieved.The identification of all brands was achieved by the second-order derivative pretreatment method，and the accurate rate was 100%.The optimal spectral pretreatment method for the non-destructive identification of soybean milk powder was second-order derivative pretreatment.