This paper presents a novel alignment method based on wander frame for SINS (strap down inertial navigation system) whose initial attitude information is completely unknown. During the course mode, the cosine and sine of the wander angle is defined as the filter states, and modeled as the velocity/position error in the geographic frame. Azimuth error
state in the fine mode of the filter is defined as wander angle error to provide continuous estimation of navigation states. A systematic analysis of the Observability of a strap down inertial navigation system (SINS) in ground alignment with Bar-Itzhack and Berman's error model is presented. It is shown that the unobservable states are separately contained in two
de-coupled subspaces. The constraints on the selection of unobservable states are discussed. An estimation algorithm, which is derived fully from the horizontal velocity outputs for computing the misalignment angles, is provided. It reveals that the azimuth error can be entirely estimated from the estimates of leveling error and leveling error rate, without using
gyro output signals explicitly. This paper also, estimates the errors of a strap down inertial navigation systems by using appropriate Kalman filter design.