Accurate vehicle orientation tracking in 3 dimensions or 3D attitude tracking, is an essential requirement for various vehicle stability and safety applications. In this paper, we consider the problem of vehicle attitude tracking along with its external acceleration estimation using low cost inertial measurement units. The current state of the art approaches are unable to estimate orientation when the body is in motion for a prolonged period, as the accelerometer measurements are corrupted by external accelerations that are produced due to body movements. In this paper, a novel filtering framework is proposed to eliminate the acceleration induced uncertainty for body observing severe and prolonged external acceleration, which is based on linear Kalman filter. We show that the sensor constraints allow the removal of external acceleration from all all of its axes. This approach is unique and superior from current state of the art as it estimates the acceleration without using any information from additional sensors such as wheel encoder, GPS, and/or camera. A unique state formulation is proposed which incorporate magnetometer to compliment accelerometer in rotation matrix frame work. Simulations and real world experimental tests are conducted to verify the performance of the proposed algorithm in various dynamic conditions.