Key Vertices based Layered Scene Maps Constructing for Campus Road Conditions

Addressing the diverse requirements of campus road condition inspection tasks in terms of map accuracy, detail, and hierarchical structure, this paper proposes a layered road scene mapping approach based on key vertices. Initially, by introducing prior information and elevation data, a global topological map with elevation is constructed based on a two-dimensional topological map, thereby reflecting the actual conditions of campus roads more precisely and comprehensively. Subsequently, the mapping of local regions at key vertices was further conducted. A local mapping method based on the CAUnet model was proposed, which integrates real-time semantic information, depth maps, and an efficient octree structure, and improves the real-time performance of map construction but also greatly enhances the map's ability to retain and distinguish environmental information and obstacles. The layered map provides solid and reliable fundamental data support for subsequent tasks such as robot navigation and path planning, potentially further enhancing the efficiency and accuracy of campus road condition inspections.