When the horizontal attitude reference of celestial navigation equipment and the star tracker are installed separately，they are affected by the temperature change of the installation structure，stress release and other factors. The direct transmission of the horizontal attitude reference to the star tracker will introduce an installation error angle，which has a significant impact on the accuracy of celestial navigation. Aiming at the estimation problem of the installation error angle between the horizontal attitude reference and the star tracker，the installation error estimation model is design and derived，the measurement component is optimized，the relative motion strategy between the measurement component and the horizontal attitude reference is designed to determine the component and the horizontal attitude reference. The difference between the three-axis gyro data between the measured component and the horizontal attitude reference is used as the measurement information of the Kalman filtering，which can realize an effective estimation of the installation error. The simulation results show that a suitable model is designed，the angular velocity of the gyro is used as the measurement information of the Kalman filtering，and the stable state of the filter can be achieved in a short time through a certain rotation strategy，and the accuracy of horizontal installation error angle estimation is better than 3.6 arcseconds，and the accuracy of gyro bias estimation is better than 0.01 degrees per hour. Therefore，the algorithm has good adaptability for the transmission of the celestial navigation level attitude reference to the measurement component，and has practical value and research significance.