Web-based enrichment of bike sensor data for automatic geo-annotation

In this paper, we describe a multi-modal bike sensing setup for automatic geo-annotation of terrain types using web-based data enrichment. The proposed road/terrain classification system is mainly based on the analysis of volunteered geographic information gathered by bikers. By using participatory accelerometer and GPS sensor data collected from cyclists' smartphones, which is enriched with data from geographic web services, the proposed system is able to distinguish between 6 different terrain types. For the classification of the web-based enriched sensor data, the system employs a random decision forest (RDF), which compared favorably for the geo-annotation task against different classification algorithms. The system classifies every instance of road (over a 5 seconds interval) and maps the results onto the user collected GPS coordinates. Finally, based on all the collected instances, we can annotate geographic maps with the terrain types and create more advanced route statistics. The accuracy of the bike sensing system is 92% for 6-class terrain classification and 97% for 2-class on-road/off-road classification.