PERCEPTUAL ARCHITECTURE FOR A VISION SYSTEM OF AUTONOMOUS VEHICLES
Volker von Holt, Stefan Baten · 1998
Early approaches to the task of perception were more or less designed as independent expert modules for specific tasks, justified by the limited complexity of scenarios. Although the results achieved with such approaches are remarkable and a superior way to evaluate recognition algorithms, the expected difficulty by merging these modules for more complex scenarios increases. The perceptual architecture for dynamic vision presented here is based on a framework, which defines a solid representation of objects. It incorporates long-term experience in the domain of vision based autonomous vehicles and is designed to handle future challenges of more complex scenarios in real time.