Abstract: In order to investigate the feasibility of fast testing the filling power of flue-cured tobacco by NIR calibration model, 94 representative flue-cured tobacco samples were collected from Henan. A model for predicting filling power was established via fitting the near infrared spectra data to the measured filling power of tobacco samples with partial least square method, the effects of the pretreatment and range of spectra on the developed model were examined, and the internal cross validation, external validation and model precision test were conducted. The results showed that: 1) The spectrum pretreatment method, which combined standard normal variable transform with first derivative, and whole spectrum range were suitable for developing the NIR model of filling power. 2) The determination coefficient of the model was 0.960 with RMSEC of 0.094, the results of internal cross validation and external validation indicated that the value predicted by the model extremely significantly correlated to the measured value. 3) The relative standard deviation of model precision test was <3%. The NIR model for predicting filling power is accurate, of good reproducibility and suitable for fast testing the filling value of batches of flue-cured tobacco.