This paper presents the helmet tracker system which can estimate the position and attitude of the pilot’s helmet even though the shape and the number of obtained features vary. The helmet tracker algorithms consist of the feature detection and 3D motion estimation algorithm. The feature detection algorithm predicts the position of the feature on the 2D image plane by using the result of the 3D motion estimation algorithm and it extracts a center point of a feature through the digital image processing technique. Then, this algorithm gets the correspondence point of the features from the both left and right image pairs. As a result, the helmet tracker can operate regardless of the helmet’s attitude and position. Finally, the 3D motion estimation algorithm using the extended Kalman filter is presented. The filter consists of nine state variables such as three positions, three velocities and three angular rate of the head frame with respect to the camera reference frame. The rotational experiment using the accurate rate table is carried out to verify the performance of the helmet tracker system. The experimental result shows that the proposed helmet tracker can extend the working volume and be operated at the dynamic environment.