Cooperative Unmanned aerial vehicles (UAVs) have been widely employed as effective tools for various information-gathering tasks in complex environments with increased efficiency and resiliency. The mission-level guidance and control of UAVs often depend on an accurate map and inaccurate maps may lead to the UAV's inappropriate accommodation to the environment. In this paper, we propose a new framework to generate and utilize semantic map information, which we defined as risk factors for cooperative UAVs. First, we generate a high-precision panorama as a global map by mosaicking a bird's-eye atlas. Afterward, we build a semantic map based on a neural network. Finally, we utilize the semantic information-enhanced map to guide the path-planning functions. Experiments show that our proposed method can improve the success rate of planning in the outdoor scene, and demonstrate its efficiency.