Abstract: Near infrared diffuse reflectance spectroscopy combined with partial least square (NIRSPLS) was applied to establishing calibration models for determining the contents of rifampicin and isoniazide in rifampicin and isoniazide tablets.The regression coefficient (R) by means of the model for determining the content of rifampicinis 0.992　77, the root mean square error of crosscalibration (RMSECV) is 0.006 65, the R by means of the model for determining the content of isoniazide is 0.989　01,  and the RMSECV is 0.00423.Using these models for predicting the contents of rifampicin and iso niazide in validation set, the root mean square error of prediction (RMSEP) are 0.005　73 and 0.003　79.These results demonstrate that this method possesses high precision of prediction. The average recoveries are 99.376% and 98.243%in the recovery experiments, and the relative standard deviation (RSD) are 0.679　3% and 0.639　8% in reappearance experiments. With its advantages of facility, rapid, nondestruction, no pollution, online detection and good reappearance, this method must be popularized in situ measurement and in the online quality control for rifampicin and isoniazide tablets production.

Key words: near infrared diffuse reflectance spectroscopy, partial least square, rifam picin, isoniazide