Fourier transform-near infrared (FT-NIR) spectra of microorganisms reflect the overall molecular composition of the sample. The spectra were specific and can serve as spectroscopic fingerprints that enable highly accurate identification of microorganisms. Bacterial powders of one yeast and five bacteria strains were prepared to collect FT-NIR spectra. FT-NIR measurements were done using a diffuse reflection-integrating sphere. Reduction of data was performed by principal component analysis (PCA) and two identification models based on linear discriminant analysis (LDA) and artificial neural network (ANN) were established to identify bacterial strains. The reproducibility of the method was proved to be excellent (Dy1y2: 1.61±1.05-10.97±6.65) and high identification accuracy was achieved in both the LDA model (Accuracy rate: 100%) and the ANN model (Average relative error: 5.75%). FT-NIR spectroscopy combined with multivariate statistical analysis (MSA) may provide a novel answer to the fields which need for rapid microbial identification and it will have great prospect in industry.