OBJECTIVE： To establish a method for the  rapid identification of Sanjiu weitai granule and quantitative analysis of moisture. METHODS： Near-infrared （NIR） spectroscopy was adopted. 38 batches of Sanjiu weitai granule were collected， NIR spectrum was determined by integrating sphere diffuse-reflectance. Conformity test model was performed by comparing consistency index （CI） value and CI limit， and quantitative model of moisture was established using partial least squares （PLS） algorithm. RESULTS： When CI value was less than 6， the established model can accurately distinguish Sanjiu weitai granule， the model was proven feasible； the correlation coefficient （r2） of quantitative model was 97.44， the root mean square error of cross validation （RMSECV） was 0.453， the root mean square error of prediction （RMSEP） was 0.748. CONCLUSIONS： The method is simple and rapid without destroying， which can be applied to fast screening of Sanjiu weitai granule in site and fast determination of moisture content.