iming at the problem that the filtering accuracy is poor or even divergent in deep integrated Global Navigation Satellite System/Inertial Navigation System(GNSS/INS) in high dynamic motion environment with strong nonlinearity and inaccurate time-varying noise statistics, an adaptive hybrid filtering algorithm is proposed. In the proposed algorithm, the hybrid filtering idea is adopted to simplify the Unscented Kalman Filter (UKF) algorithm. According to the high dynamic system measurement noise time change, especially easy to change quickly and abrupt, an adaptive measurement noise estimator based on fading memory exponent is designed. It estimates and corrects the statistical characteristics in real time, and adaptively regulates the estimation cycle. The simulation results show that, in the case of variation of measurement noise, the accuracy of the algorithm is raised in comparison with the conventional UKF algorithm. The improvement effect of horizontal direction precision is obvious, which is more than 60%. In addition, the time consumption is reduced by 18.64% compared to the conventional UKF algorithm.