In a large maneuvering scene, the location estimation error and the replenishment frequency of stars occur. To solve these problems, a star tracking algorithm assisted by gyroscope is proposed in this work. In a large maneuvering scene where the angular acceleration of the carrier changes dramatically, the angular velocity information measured by gyroscope is invested to keep track of the identified stars. Thus, the center-star-points of two-consecutive frames are easily obtained. The motion model of star points is established, and the position of the star points are tracked and predicted by Kalman filter. The positions of stars inside and outside the star map can be predicted in advance. The cache star database proposed in this paper can realize fast star replenishment in a large maneuvering scene. Experiments confirm that the error of star tracking is less than 2-pixels. Furthermore, the cache database reduces time of star replenishment. Both results prove that the proposed algorithm is effective, and conduces to improve the dynamic performance and update rate of star sensor.