Because of the problem that the airborne star tracker for celestial navigation is affected by the external environment easily, which leads to performance degradation of celestial navigation, an anomaly detection algorithm is proposed. The error model of observation information of star tracker relative to the theoretical observation based on inertial navigation information is established, and the field view threshold detection, 3? division and slope estimation detection are used to detect the outliers, jumps and slow changes in the observation information respectively. The experimental results show that this method can detect the anomaly information of celestial observation effectively, which lays a foundation for the engineering realization of airborne CNS/INS integration navigation system by use of star tracker.