In order to meet the requirements of real-time planning of safe obstacle avoidance trajectory for intelligent vehicles in dynamic scenes, a stable obstacle avoidance trajectory planning method for intelligent vehicles based on double Bezier curves is proposed. The traditional Bezier curve is improved, and the trajectory planning strategy of double Bezier curve with transit state is established, and the transit position is optimized by the minimum longitudinal safety distance model to ensure the safety of obstacle avoidance. Combining with the scene information, the vehicle dynamic constraint optimization obstacle avoidance strategy is established to ensure the stability of the obstacle avoidance process, so as to obtain a reasonable trajectory set. An evaluation function is designed based on safety and stability to obtain an optimal obstacle avoidance trajectory with comprehensive performance from a reasonable trajectory set. CarSim/Simulink co-simulation and hardware-in-the-loop experiment are carried out. The experimental results show that the intelligent vehicle keeps a certain safe distance from the vehicle in the process of obstacle avoidance, and the maximum yaw rate, lateral acceleration and roll angle of the vehicle are 10.1 (°)/s, 2.1 m/s2 and 0.91°, respectively, which proves that the proposed trajectory planning method can generate the optimal obstacle avoidance trajectory under various working conditions, and can ensure that safe and stable completion of the obstacle avoidance proces.