The position and attitude accuracy of the FAST (The Five-hundred-meter Aperture Spherical radio Telescope) feed cabin directly affects the alignment accuracy of telescope's receiver. In order to improve the position and attitude accuracy of the feed cabin, this paper proposes the attitude Kalman filtering algorithm. The attitude of the feed cabin is solved by BDS (BeiDou Navigation Satellite System) multi-antenna technique, and the attitude difference between BDS and SINS (Strap-down Inertial Navigation System) is taken as the measurement value of the Kalman filter. In order to reduce the complexity of solving BDS attitude, the lodrigues matrix is used to solve this attitude. In this paper, the actual measurement values of the BDS and SINS are taken as the test data, when the telescope is in tracking mode. The test results show that the accuracy of the integrated navigation using attitude Kalman filtering is better than that of the traditional Kalman filtering, especially the accuracy of the yaw angle and y_f axis position.