When using airborne optoelectronic platforms for “air-to-air” target tracking, the tracking effect is always affected by the environment and it is easy to lose the completely-sheltered target.Based on the traditional CamShift algorithm, a dynamic target tracking and trajectory prediction algorithm based on multi-feature fusion and relative Kalman model is proposed.In target template construction, the method of fusing the color, texture and gradient features is adopted to improve model description.During the tracking period, the dynamic update of feature templates is introduced to ensure the long-term stability of the algorithm.In the case of complete occlusion, the Kalman models are established respectively for background center and the difference between the flight target and the center point through the secondary occlusion judgment method, which greatly decreases the probability of misjudgment and frame loss.Experimental results show that the proposed algorithm has high accuracy, fine real-time performance and stability.