In this paper， using a Fourier transform near-infrared analyzer and an ultraviolet-visible spectrophotometer， with a commercial active beverage as a research object， a rapid quantitative detection study on the number of live lactic acid bacteria in live bacterial beverage was carried out. The original spectra of live bacterial beverages were subjected to Savitzky-Golay （SG） smoothing combined with first-order derivation （FD）， multivariate scattering correction （MSC） and orthogonal signal correction（OSC） pretreatment， and then using partial least squares to establish lactic acid bacteria a quantitative prediction model for the number of viable bacteria， and optimization and external verification of the optimal model. For the number of live lactic acid bacteria in live bacterial beverages， the PLS model established by Fourier transform near infrared spectroscopy with OSC pretreatment was the best. Using competitive adaptive weighting method （CARS） for wavelength screening and model optimization， the correction set correlation coefficient and root mean square error Rc was 0.9974， the RMSEC was 0.0211， the verification set correlation coefficient Rp was 0.9837， the root mean square error RMSEP was 0.0508. Finally， using the commercial samples that did not participate in the modeling， the optimized Fourier transform near infrared prediction model was externally verified. The correlation coefficient of the predicted value was 0.9068， and the root mean square error was 0.0108. The results showed that based on Fourier transform near infrared spectroscopy， the number of lactic acid bacteria in live bacterial beverages could be quickly and accurately detected， which provided a quick and convenient method for quality monitoring in the production and sales of live bacterial beverages.