Outlier detection and handle in transfer alignment

Since the accuracy of the Kalman filtering estimation would be seriously decreased or even divergence during the transfer alignment when datum information contains outliers. Based on the analyzing the principle of outliers affecting Kalman filter, a method of detecting and inhibiting outliers is proposed. This method uses the absolute value of innovation and the ratio of trace of innovation actual variance and theoretical variance to detect outlier. When outliers are detected, gain matrix is reconstructed to inhibit the effect of outliers on the filter. At last, it is verified by the transfer alignment simulation system. The results show that the method can effectively inhibit the effect of outliers on the Kalman filtering estimation and improve filtering performance.