Main Entrance Tagging based on Intrinsic and Extrinsic Feature Extracted from OSM

Abstract

Knowing the location of the main entrance is significant in accurate navigation. Current LBS providers, such as Google Map often guide users to a wrong location that is far away from the main entrance. This is an unpleasant experience especially for the mobility-impaired people. Realizing the importance of the entrance information, many buildings on OpenStreetMap (OSM) have been tagged with the entrance. However, the proportion is still small. To mitigate this gap, this work proposes using machine learning to inferring the location of the main entrance of a building based on its spatial contexts (e.g., main road) and its footprint (e.g., centroid).