Among many application fields of unmanned aerial vehicles (UAVs), completing the exploration of unknown areas is a fundamental problem. In this paper, we propose a method to introduce autonomous return into UAV exploration, providing a solution for UAV exploration in a large-scale unknown environment. We first build the system architecture of UAV exploration and outline the workflow by frontier information module and path planning module. Among them, in the frontier information module, we introduce a construction method for frontier viewpoints, and each viewpoint can maximize the coverage of frontier information. Further, through the modeling of the Traveling Salesman Problem (TSP), the optimal global path of the UAV covering the viewpoint is obtained. Then the path planning of autonomous return is added to the process of UAV exploration, and the return task is completed through kinodynamic path searching. Finally, this paper examines the coverage rate of the proposed method in the case of limited exploration time and verifies the feasibility of the proposed method.