Initial alignment of strap-down inertial navigation system (SINS) is nonlinear when the azimuth misalignment is large angle.The error equations contain linear state and nonlinear state.The estimation of single linear filter algorithm is not optimal.In order to improve the initial alignment accuracy and shorten the initial alignment time,the mixed dynamic filter which is Unscented Kalman filter- Kalman filter (UKF-KF) is used in initial alignment of large azimuth misalignment angle.The azimuth misalignment angle is separated from the horizontal misalignment angles.The Unscented Kalman filter is used to estimate the azimuth misalignment angle.The Kalman filter is used to estimate the horizontal misalignment angles.Then it takes the estimation error of azimuth misalignment angle to correct the horizontal misalignment angles.Simulation results show that compared with the Unscented Kalman filter the Unscented Kalman filter- Kalman filter improves the estimation accuracy and shortens the estimation time.