In this paper, a method based on neighborhood cognitive consistency is proposed for the UAV cluster cooperative target rounding problem. Firstly, according to the task characteristics of target roundup, the corresponding mathematical model and roundup success determination rules are established; secondly, according to the characteristics of individual UAVs in the cluster, graph convolutional network is used to represent the information interaction relationship between UAVs, and through the neighborhood cognitive consistency method, the consistency of the neighborhood UAVs with respect to the target information is achieved, which promotes the UAVs to carry out the collaboration; lastly, through the design of the reward function, it is promoted that the UAV clusters to collaborate to complete the roundup task. Simulation experiments show that the method adopted in this paper can successfully avoid obstacles and smoothly complete the roundup task, and has a higher roundup success rate and convergence speed.