Towards Local Dynamic Maps: Automatic Validation and Correction of Road Map Features

Abstract

The creation and maintaining of traffic sign databases requires a lot of manual effort. On the one hand, those databases are necessary for various use-cases including landmark-based localization, risk assessment, traffic management / speeding ticket. On the other hand, those databases require updates on a regular basis because of changes in the real world, e. g., due to vandalism, construction work, or accidents. For these reasons, databases are usually not up-to-date and contain a lot of inconsistencies w.r.t. the real world. Our goal is to automate the process of updating traffic sign databases using mobile sensor platforms. Our solution is to verify and correct the database on the fly leveraging the vehicle's ability of localization and perception. To that end, we set up a complex modular processing pipeline