To solve the tracking error caused by camera image distortion, motion blurring, air drag and other uncertain factors, a tracking algorithm based on air “drag factor” estimation was presented in this paper. This algorithm incorporated the tracking advantage of Kalman filter, added a “drag factor” term which was deduced from aerodynamics theory to state vector, and re-modeled the augmented non-linear state equations. Finally, a non-linear extend Kalman filter was applied to estimate the model’s states. The experimental results indicate that the improved method has more precise output than traditional tracking algorithm (velocity noise reduced from 0.5 m/s to 0.2 m/s). Its efficiency and low calculation load in computation was also proved.