Aiming at the problem of insufficient real-time and reliability of UAV path planning in unknown complex environments, a path planning algorithm based on heuristic graph search is proposed. A memory-based local navigation histogram is constructed to represent the environment, and an objective function that adapts to the flight characteristics and mission requirements of the UAV is designed. An adaptive weight term is introduced to adjust the objective function according to the flight state. The algorithm is verified in the Gazebo simulation environment, and the experiment shows that the algorithm can plan a reasonable path that meets the constraints in real time, improving the flight quality.