Abstract: Dual quaternion is applied to the description of strapdown inertial navigation system (SINS), which simplifies the initial alignment model and reduces computational complexity by considering the rotation and translation of coordinate system together. A dual quaternion based coarse alignment model of SINS initial alignment is given, and two extractive accurate alignment models are also presented here in terms of additive and multiplicative dual quaternion error model. Taking yaw angle estimation as an example, simulation results show that algorithm based on dual quaternion could save 40% of convergent time with an improvement of 0.01° to alignment precision against traditional method.