Leaf area index (LAI) can be used as a monitoring index for assessing tobacco health in different growing periods. Hence, acquiring and updating accurate LAI data in a timely manner are necessary for managing the growth of tobacco. Growth information for tobacco in different seasons could provide valuable information for management on a national scale. Accurate and continuous tobacco LAI dynamics data are based on the data fusion framework of the ensemble Kalman filter (EnKF), which is an efficient recursive filter to estimate the state of a dynamic system from a series of incomplete and noise measurements, can be used to obtain optimal results. Nanxiong City in Guangdong Province was selected as the study area to extract LAI data for tobacco and test the effect of EnKF method on the basis of quantitative remote sensing data for the growing state of tobacco in 2014. Tobacco canopy hyperspectral reflectance data in different growing seasons were collected every 15 days by using AvaSpec-ULS2048 HandHeld spectroradiometer made by Avantes company in the Netherlands. Tobacco LAI data were retrieved using the Normalized Difference Vegetation Index (NDVI), which was calculated using the reflectance data. An improved tobacco growth model (LOGISTIC) was established using the LAI data collected around Nanxiong. This improved model used LAI and accumulated temperature to reveal the changes in LAI in different growing seasons. On the basis of integration of LAI data (obtained using remote sensing data) and LAI data (obtained using the simplified LOGISTIC model and EnKF method), continuous LAI data were obtained in the time series during the tobacco growing season in Nanxiong. Finally, we compared three different LAI computing methods in tobacco study: (a) calculated by NDVI, (b) simulated by the LOGISTIC model and (c) data assimilation was based on EnKF. The results indicated that these three methods could describe the growth status of tobacco to a certain extent, especially in the mature growth period, however, the LAI assimilation method was the best, which was able to adjust measured values and model values dynamically, LAI data were more consistent with the practical growth conditions of tobacco. Method (a) was imperfect at early and late growth seasons of tobacco (in these two seasons, the LAI data were either less or more), and method (b) was more dependent on accumulated temperature data (LOGISTIC model could not effectively describe the unexpected changes in tobacco LAI). The results showed that the EnKF algorithm could obtain better estimation on the basis of the dynamic model, and assimilate remote sensing data into the dynamic model to obtain optimal estimation for LAI. The assimilated LAI data were closer to the real values, and the LAI curve was more consistent with actual tobacco growth status.