Ultraviolet ( UV ) spectrum method was used to measure the spectra of flue-cured tobacco, and the spectra of different grades of flue-cured tobacco were compared. The samples were scanned in 210-360 nm with UV spectrometer. Further, the UV predicting models of total nitrogen of tobacco leaves were established by stepwise multiple regression method, principal components analytic method and partial least squares method. The results show that: (1) Predicting models of total nitrogen were established to adopt stepwise multiple regression method and principal components analytic method. (2) The models of total nitrogen of the different grade flue-cured tobacco established through scanning within the wavelength range of 230-290 nm were better for total nitrogen prediction. (3) The accuracy of total nitrogen prediction for flue-cured tobacco leaf with principal components analytic method was better than that of stepwise multiple regression method, and that of principal components analytic method is 78%. (4) The prediction model is different with different data treatment method. (5) The accuracy of prediction model could be improved by adopting fitting characteristic spectrum curve of total nitrogen. It is shown that the model of total nitrogen content prediction for flue-cured tobacco can be established by using UV spectrum. The model features good stability and can be used to determine accurately and speedily the total nitrogen content of flue-cured tobacco leaves without pollution.