Versatile Geo-referenced Maps for Indoor Navigation of Pedestrians

We propose a system enabling users to capture the data needed to construct maps of building interiors, complemented by a smartphone-based pedestrian navigation application utilizing these maps. For this purpose, the user captures photographs of evacuation plans which are processed by the system using a suitable image processing pipeline. As a result, a coarse floor map is available, consisting of polygonal room outlines and identified staircases, where possible. Furthermore, the system derives the user’s initial position (which is marked in the plan) and orientation from the photographed plan. The coarse model can be further refined by geometric or semantic data geo-referenced by a pedestrian positioning solution. In a prototype configuration, two foot-mounted IMUs are used to process position and orientation with an extended Kalman Filter, using the such derived positions of features as door openings, stairs and elevators. In our system, the user downloads these traces from a repository. We make the final floor map available to the user. The coarse area-based model is transformed into walkable paths, annotated with additional semantic annotations, as available on the evacuation plan. Following this, smartphone users can easily use these maps for indoor navigation using only locally available information, without the need for a the building owner to make the maps available to the general public. Our proposed system therefore allows the easy creation and refinement of indoor maps directly applicable for navigation. Keywords—foot-mounted IMU; indoor navigation; floor maps; map refinement