Aiming at the problem of autonomous obstacle avoidance of UAV in complex dynamic environment, an improved speed obstacle method is proposed to improve the autonomous obstacle avoidance ability of UAV in complex environment. Firstly, according to the principle of velocity obstacle method, a mathematical model is constructed to determine whether there is conflict between UAV and dynamic obstacle. Then, the adaptive safety distance is added and the Kalman filter is used to predict the position of the dynamic obstacle at the next moment, and a velocity obstacle area is reconstructed. Finally, an improved dynamic window method is used to solve the optimal obstacle avoidance strategy for the feasible obstacle avoidance region. The feasibility and effectiveness of the proposed algorithm are verified by the simulation experiments of different scenes and multiple obstacles.