The detection and alarm of air threat situation is the key to the early warning for UAV flight collision avoidance. For this reason, a detection and alarm method of air threat situation based on track prediction is proposed for UAV collision avoidance. In the method, firstly, the sliding window polynomial fitting method is used to dynamically predict the track of the intruder. Then, on the basis of the track prediction, using the flight information of UAV and intruder, the dynamic collision avoidance zone of UAV is established to the intruder. Finally, combined with static protection zone principle, the detection and alarm for flight conflict trend, nearSmidairScollision trend and flight collision trend between UAV and intruder is carried out in a specific flight scenario. Simulation results validate the effectiveness of the proposed method and the feasibility of application in the early warning for UAV flight collision avoidance.