Kopplung von Sensorinformationen und Geodaten zur modellbasierten Lokalisierung und Kartierung im Automobil (Volume 51) (English shop)

Preview

Table of Contents, Datei (49 KB)
Extract, Datei (190 KB)

The ongoing development of driver assistance and safety systems continually generates new demands on environment perception and the creation of an environment model. Urban areas, with their high traffic and environmental dynamics, pose a particular challenge for sensory perception. In order to avoid accidents or reduce the risk of injury, for example at complex intersection scenarios, vehicle-based environment perception must be extended to encompassing concepts. Such approaches depend both on precise vehicle localisation and on a description of the surroundings that is as complete as possible. For automotive applications, the focus is less on integrating expensive additional sensors than primarily on the intelligent combination of already proven ones. The camera, as a sensor with a high information density, is enjoying an increasingly versatile range of uses.

This work presents an image-based approach for the precise position determination of a vehicle with simultaneous mapping in urban environments. The focus lies on buildings and in particular on their salient vertical edges, which are detected by a camera and reconstructed in 3D. Matching these features with an automotive-compatible base map enables the stepwise refinement of the vehicle pose. At the same time, the additional information generated is stored in local maps, from which conclusions about structural changes in the base material are subsequently drawn. This creates a cyclically complementary form of description for the vehicle environment.