To solve the time-consuming problem of traditional detective methods,such as total viable counts and ATP bioluminescence technique,and the complicated operations of the existing spectrum techniques,a method for the quantitative analysis of the total number of bacteria in milk by UV-Vis spectroscopy combined with chemometrics,followed by application of the optimized partial least squares to set up the model of the total amount of milk bacteria.With the milk cultured in the same time period(0，2，4，…，20 h) as target,the quantitative relationship between the spectrum and the total number of milk bacteria was explored.The successive projections algorithm was applied to select 19 characteristic waves,and then the partial least squares model was used to establish the model relationship between the characteristic waves and the total number of milk flora.The correlation coefficient R of 0.978 in the optimized model was higher than that of 0.968 in primitive model,and the root mean square error(RMSE) of 0.265 in the optimized model was lower than that of 0.324 in the primitive model.The proposed method provided a new idea and direction for the rapid detection of the number of milk flora.