This article presents a novel approach to identifying and classifying intersections for semantic and topological mapping. More specifically, the proposed novel approach has the merit of generating a semantically meaningful map containing intersections, pathways, dead ends, and pathways leading to unexplored frontiers. Furthermore, the resulting semantic map can be used to generate a sparse topological map representation, that can be utilized by robots for global navigation. The proposed solution also introduces a built-in filtering to handle noises in the environment, to remove openings in the map that the robot cannot pass, and to remove small objects to optimize and simplify the overall mapping results. The efficacy of the proposed semantic and topological mapping method is demonstrated over a map of an indoor structured environment that is built from experimental data. The proposed framework, when compared with similar state-of-the-art topological mapping solutions, is able to produce a map with up to 89% fewer nodes than the next best solution.

翻译：本文提出了一种用于语义与拓扑建图中识别和分类路口的新方法。具体而言，该方法能够生成包含路口、路径、死胡同以及通往未探索边界路径的语义地图。此外，生成的语义图还可用于构建稀疏拓扑地图，供机器人实现全局导航。该方案引入了内置滤波机制，既能处理环境中的噪声，又能移除地图中机器人无法通行的开口，并剔除小型物体以优化和简化整体建图结果。通过基于实验数据构建的室内结构化环境地图，验证了所提语义与拓扑建图方法的有效性。与现有同类顶尖拓扑建图方案相比，该框架生成的地图节点数最多可减少89%。