Aiming at the problem of uncertainty of GNSS measurement noise variance parameter in the traditional combined navigation filtering algorithm, this paper, based on the principle of SINS/GNSS dynamic differential sequence, improves the traditional Sage-Husa adaptive extended Kalman filtering algorithm (AEKF) method of estimating the measurement variance array based on the information of residual sequences, utilizes the high-precision characteristics of short-term positioning of SINS and combines with the smoothing of bounded layers to isolate the abnormal observation information of GNSS, so that the improved adaptive filtering algorithm can maintain a high level of positioning accuracy under different noise environments of GNSS. The fault detection algorithm isolates the abnormal observation information of GNSS, so that the improved adaptive filtering algorithm can maintain high positioning accuracy under different noise environments of GNSS. The experimental results of the actual sports car show that in the low-density anomaly noise environment in GNSS work, the algorithm in this paper improves the average positioning accuracy by 39.9% and 7.9% compared with the EKF algorithm and the traditional Sage-Husa algorithm, and in the high-density anomaly environment, the overall positioning accuracy is improved by 64.5% and 29.1%. Therefore, the algorithm in this paper effectively improves the anti-interference ability of the combined navigation system against different measurement noises.