In order to solve the inaccurateness of nonlinear error model of inertial navigation system(INS) transfer alignment
with large azimuth misalignment, tow major factors are taken into account, which are flexure and lever-arm effect. An
incorporate acceleration error model of these factors is built, which make the nonlinear error model of transfer alignment
accurate. In view of stability and rapidity of nonlinear filter, the scaled correctional unscented Kalman filter(UKF) is used
to estimate attitude misalignment. The velocity matching is used to validate the inaccurateness of nonlinear error model and
validity of nonlinear filter. The simulation results show that this model can meet the precision and time request in transfer
alignment with large azimuth misalignment.