An improved unmanned aerial vehicle (UAV) low altitude penetration model based on safe flight space is presented to solve the problem of lack of combination with the characteristics of UAV after terrain preprocessing. A safe flight space is constructed to reduce terrain constraints and ensure the safe flight of UAV in low altitude penetration. The trajectory planning model is improved by combining with the characteristics of UAV to increase the stability of UAV flight process and the smoothness of planned trajectory. The particle swarm optimization (PSO) algorithm is used to simulate the two models before and after the improvement in complex elevation map to verify the rationality and effectiveness of the improved model.