In order to predict the impact point precisely and instantly, the state equation of Kalman filtering, based on the centroid ballistic model, has been built in the paper. By making using of memory attenuated extended Kalman filtering, unknown parameters and wind de-flection in the state equation were estimated, which were used to extrapolate the impact point. As the result of checking with the real data of satellite seeker, the drag acceleration coefficient could be converged rapidly and the influence of wind deflection could also be simulated. What's more, the kalman filtering algorithm for impact point prediction has less computational cost and is more accurate and simpler for application in engineering.