In recent years, advanced technologies such as sensing, data analysis and intelligent control have promoted the reform in the field of disaster rescue, and the application of UAVs has become more and more widely based on their advantages. In related research, the classical ant colony algorithm, particle swarm optimization algorithm and the emerging artificial bee colony algorithm, firefly algorithm are widely used in the cooperative control of UAV swarm. In this paper, we propose a dynamic path planning method for multiple UAVs, covering path planning, target recognition, and intelligent obstacle avoidance. Path planning realizes task planning with different decision intentions according to different weight coefficients. The target recognition was transmitted to the YOLOv5 network through the UAV image to identify the target and calculate the position. Intelligent obstacle avoidance searches path based on A* algorithm. Through simulation experiments, the algorithm is tested in static and dynamic situations, which verifies its accuracy, stability and efficiency in initial path planning and intelligent obstacle avoidance.