Validation of HD maps for autonomous driving

Abstract

This contribution describes a new way of generating highly accurate high definition (HD) maps of the road infrastructure and the necessary validation process ensuring high geolocation demands. As data source aerial imagery acquired with a camera system onboard a helicopter are used. A detailed description outlining (a) Data Acquisition, (b) Orthorectification and (c) Combination with highly accurate radar measurements performed by the TerraSAR-X satellite system is given. The highly accurate imagery geolocation is validated using Ground Control Points (GCP) acquired independently within a GNSS measurement campaign, a description of the final absolute geolocation accuracy is given. A HD map is created using the refined aerial image consisting of road markings (e.g. solid/dashed) and guard rails, which can serve as reference map for validating HD maps generated with other systems.