GPS/INS integrated systems are applied in the self-positioning of the high dynamic targets. This paper forms state equations and observation equations by the integrated system and GPS observation respectively. An adaptive filtering algorithm is introduced which estimates observation noise covariance with observation data, and estimates dynamic noise covariance by use of Sage-Husa filtering algorithm which has faster convergent speed and better stability than the conventional Kalman filtering algorithm. At last, simulations are given to verify the effectiveness of the algorithm.