Analysis and simulation results indicate that two problems should be solved when the self-alignment method based on gravitational apparent motion and dual-vector can be used for Strapdown Inertial Navigation System. The first one is how to identify the apparent motion from accelerometer measurement containing random noise and the second is how to avoid the collinear problem between two vectors used in alignment solution. In this paper, a parameter identification and reconstruction algorithm is proposed to solve the first problem and simulation results indicate that proposed algorithm can identify apparent motion from accelerometer measurements effectively; and reconstruction algorithm based on current identified parameters for dual-vector is designed in detail to solve the second problem which can make full use of newest identification and avoid collinear problem completely. Simulation and turntable results show that the proposed alignment method can fulfill self-alignment in a swinging condition and the alignment accuracy can reach the theoretical values determined by the sensor precision.