Attention control integrated in a system to autonomous driving and collision avoidance under egomotion

In case of vision based driver assistance a processing in real time is necessary. In analogy to the visual system in humans this can be achieved by focussing the limited resources of a system to those parts of an image that are relevant to the task. Within this framework we present an approach to attention control that is embedded in a system for traffic scene analysis (MOSAIK). Based on motion analysis and strategies for perceptual grouping regions of interests, so-called attention fields, are detected. The integration of such attention mechanisms yields the advantage that a detailed recognition is only necessary on selected parts of an image. We show that the interaction between supervising attention control, detailed recognition and fast tracking supports a very efficient analysis. So MOSAIK computes a robust scene description nearly in real time and that on complex scenes, too