The estimation of flight state for guided projectiles is one of the most critical techniques in obtaining precise trajectory parameters. When Kalman filtering is used, the system model deviation may cause reduction of the filtering accuracy significantly. In this paper the foregoing problem is investigated by taking a type of fin-stabilized projectiles with low spin speed as object. An Extended Point-Mass Model (EPMM) is proposed. Based on the EPMM and radar detecting system, the state equation and measurement equation are established respectively. The effect of model deviation on the filtering performance is analyzed via numerical simulation, thus a new scheme of estimation of flight state is proposed based on the process noise control. The results indicate that the negative influence is overcome effectively by adopting this scheme. The estimation error of lateral range for projectiles can be reduced more than 50% at least and the increase of computation burden is fairly small. Some results concluded are useful for further research.