A graph-based topological maps generation method for indoor localization

Indoor maps are widely used to display user's location and refine pedestrian trajectories by enforcing constraints such as impassable walls. However, indoor maps are always unavailable due to their time-consuming and labor-intensive manual constructions. The widespread CAD drawings enable us to generate indoor maps at affordable costs. In this paper, we present a graph-based method for automatically generating topological indoor maps. We preprocess the initial data from CAD drawings and extract qualified door lines and wall lines. Rooms and corridor are extracted by detecting minimum cycle basis (MCB) of a walls-based graph. The indoor maps are then constructed by analyzing topology between all indoor spatial elements. In order to validate the generated maps, map matching algorithm using particle filter is implemented to calibrate the preliminarily estimated trajectories. Experiments results show that the proposed method runs much faster than previous work and the generated maps can significantly improve the average accuracy. Meanwhile, the number of cross-wall behaviors is also reduced.