Abstract: To solve the problem of poor state estimation accuracy of ballistic correction projectiles in ballistic environment, the data during the experimentation was utilized to separate the ballistic measurement error sequence during the projectile’s flight from the satellite positioning and velocity measurement data of the projectile and the radar data. The error probability distribution was approximately fitted using the Gaussian mixture model (GMM), the expression form of which is unified. The traditional Gaussian mixture extended Kalman filter (GMEKF) algorithm was improved by considering the correlation of the noise among multiple adjacent moments. The colored observation noise was decoupled using the AR model and whitened using the state augmentation method and the difference method. Taking the trajectory simulation of the correction projectile as an example for algorithm verification and comparison, the experimental results demonstrate the effectiveness of the improved GMEKF algorithm in improving the estimation accuracy of ballistic parameters and the accuracy of the impact point.