Abstract: To determine the chlorophyll content in tobacco leaf real-timely, nondestructively and rapidly, the spectrum of healthy leaves of different vigor of main tobacco cv. Nanjiang No.3grown at karst mountain area in Guizhou were collected by a Field Spec 3 portable spectrometer, and the chlorophyll contents in corresponding leaves were determined. The spectral characteristic variables were extracted via spectral analysis, and the correlations of chlorophyll content in tobacco with the original spectrum, first derivative of spectrum and spectral characteristic variables were analyzed. An estimation model for chlorophyll content was established by multivariate stepwise regression and the single variable of spectral characteristics, and theaccuracy of the model was tested. The results showed that: the correlation coefficients between chlorophyll a, b and original spectral reflectance peaked at 700 and 701 nm, respectively; those between chlorophyll a, b and spectral derivative reflectance peaked at 623 and 653 nm, respectively; the correlation coefficient between chlorophyll a and spectral characteristic variable λg and that between chlorophyll b and SDr/SDb were the highest. The stepwise regression model based on first derivative spectrum reflectance features high accuracy in estimating the contents of chlorophyll a and b in tobacco leaves.