VISION-BASED DRIVER ASSISTANCE USING RANGE IMAGERY
Karin Sobottka, Horst Bunke · 1998
Our world is three-dimensional, considering time as additional dimension, even four-dimensional. Despite most approaches in vision-based driver assistance have utilized greylevel or color image sequences. Since the spatial arrangement of scene objects is often more relevant than the re#ected brightness information, there has been an increasing interest in range sensors for collision avoidance systems recently. This paper presents an approach to obstacle tracking in range image sequences of tra#c scenes. Our tracking scheme uses a feature-based approach. To #nd correspondences between features in consecutive frames we strongly make use of the three-dimensional sensory information. Thus we are also able to handle the most general case in matching, the m:n matching. Alternative feature assignments are kept for selecting the most optimal consistent subset of assignments. Further temporally global processes such as aging, the prediction of 3D positions and merge of similar moving obstacles ...