Abstract: The relationship between the make-up of cigarette filler and NIR spectra was studied.Four recognition models for predicting the contents of cut stem, expanded tobacco, and reconstituted tobacco in different cigarette blends were developed by analyzing the NIR spectra of 482 tobacco samples with partial least square method.The results showed that the veracity of the predictive model for blend containing cut stem and cut lamina was good, the average square root error and average relative error of samples from different batches were below 0.9% and 7.6%, respectively;while the veracity of the predictive model for blend containing cut stem, expanded tobacco, reconstituted tobacco, and cut lamina was not as good, the average relative errors were around 10%.