Accuracy of traditional initial alignment on static base for launch vehicle strapdown inertial navigation system (SINS) depends on gyro and accelerometer bias. Under wind disturbance, error of traditional initial alignment is larger. Two-position initial alignment was used to improve accuracy of initial alignment for launch vehicle SINS under wind disturbance. The multiple model adaptive Kalman filter was presented to get optimal estimation for two-position initial alignment under the disturbance. The simulation results show accuracy of two-position initial alignment is better than traditional initial alignment, and multiple model adaptive Kalman filter can efficiently improve accuracy of initial alignment under wind disturbance.