Abstract: The estimation model of soil total nitrogen and alkali hydrolysable nitrogen was constructed based on hyperspectral data, which might contribute a new method for accurate and rapid detection of total nitrogen and alkali hydrolysable nitrogen in tobacco growing soil. Soils were sampled from Huidong and Huili, Sichuan Province, and the soil spectral reflectance data were obtained by hyperspectral imaging technique. The successive projection algorithm (SPA) and correlation analysis (CA) were employed to screen feature band, while partial least square regression (PLSR), ridge regression (RR) and kernel ridge regression (KRR) models were constructed to estimate the contents of total nitrogen and alkali-hydrolyzed nitrogen in soil by using whole and feature band, respectively. Results showed as the followings. 1) The accuracy of the estimation model was enhanced after the original spectrum was processed by four preprocessing methods. After the first derivative (D1) combined with the standard normal variate (SNV), the estimation models of total nitrogen and alkali hydrolyzed nitrogen contents established by using whole band exhibited high accuracy. 2) By using SPA, 10 feature bands of soil total nitrogen and 13 feature bands of soil alkali-hydrolyzed nitrogen were screened out, accounting for 2.58% and 1.98% of the total bands, respectively. 3) After the original spectrum was processed by D1-SNV, better performance was found in the KRR estimation model of total nitrogen and alkali hydrolyzed nitrogen content constructed by SPA screening feature bands. The coefficient of determination (R²), root mean square error of validation (RMSEV) and residual prediction deviation (RPD) of the validation set of total nitrogen estimation model were 0.87, 0.23 and 2.77 respectively. The R², RMSEV and RPD of the validation set of alkali hydrolysable nitrogen estimation model were 0.91, 14.15 and 3.39 respectively. For tobacco growing soils in the research area, the model constructed by SPA and KRR can estimate the total nitrogen and alkali hydrolyzed nitrogen contents, whereas D1-SNV-SPA-KRR method can achieve accurate estimation on the contents of total nitrogen and alkali hydrolyzed nitrogen.