Drivable space detection or road perception is one of the most important tasks for autonomous driving. Sensorbased vision/laser systems may have limited performance in bad illumination/weather conditions, a prior knowledge of the road from the map data is expected to improve the effectiveness. This paper is to employ the map information extracted from the OpenStreetMap (OSM) data, and explore its capability for road perception. The OSM data can be used to render virtual street views, and further refined to provide the prior road mask. The OSM masks can be also combine with image processing and Lidar point clouding approaches to characterize the drivable space. Using a Fully Convolutional Neural Network (FCNN), the OSM availability for deep learning methods is also discussed.
Exploring OpenStreetMap Capability for Road Perception
2018 IEEE Intelligent Vehicles Symposium (IV) ; 1438-1443
2018-06-01
1526338 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
EXPLORING OPENSTREETMAP CAPABILITY FOR ROAD PERCEPTION
British Library Conference Proceedings | 2018
|Evaluating the Capability of OpenStreetMap for Estimating Vehicle Localization Error
BASE | 2019
|Exploring OpenStreetMap Publicly Available Information for Autonomous Robot Navigation
Springer Verlag | 2015
|