A large misalignment error model of strapdown inertial navigation system(SINS) was deduced for the large relative attitude of rapid transfer alignment existing between the master and slave SINS. The Euler angles were introduced to describe the attitude errors. The Euler's kinematics equations were applied to achieve the accuracy error model. The unscented Kalman filter (UKF) was applied to avoid the derivation calculus of the attitude measurement equations which were complex nonlinear functions. The singular value decomposition (SVD) was used to solve the ill-conditioned problem of covariance matrix. The simulation experiment results show that the alignment accuracy is improved both in small angle uncertainties, and large angle uncertainties.