Aiming at the problem that the tracking data of visual head posture tracker involve time-varying noise, a method for cockpit head posture tracking based on data fusion is designed. According to the corresponding relationship between the space points and image points, orthogonal iteration algorithm is used at first to solve PNP problem. Then, the problem of the time-varying noise is solved by using Sage-Husa adaptive filter instead of EKF. Finally, a linear weighted fusion algorithm is used to fuse head posture data of the Outside-in tracker and Inside-out tracker to obtain head posture with more stable accuracy. Empirical data are used in a series of simulation, the results show that the tracking accuracy of the system is greatly improved and the system is stable and reliable.