Schmid 1994: A model-based approach for 3D- object recognition in the context of autonomous road vehicle guidance has is presented by interpreting monocular image sequences of motorway scenes in real time, even in case of partial conclusion. The image sequences are taken by a camera mounted behind the windscreen of a vehicle moving relative to the real road environment. The processing of the three following vehicle classes has been demonstrated: trucks or buses, vans and hatchback cars. The adaptation of the unknown real form parameters within each object class results automatically during run time from the simultaneous estimation of shape and motion in 3D. Because of the generic internal 3D shape model new vehicle classes could be introduced quite easily. The approach described is based on dynamical models for the motion and on geometrical generic 3D shape models of the objects expected in the motorway scenery. This so-called 4D approach (space and time) is extended to a system architecture by knowledge-based methods to assess the hypotheses derived from the video images. The internal models are updated with the optical measurements of edge features extracted from actual images by recursive estimation (extended sequential Kalman-Filter implementation). This way all the modeled 3D state variables and 3D shape parameters can be estimated in real time simultaneously. All steps of object recognition like 2D detection, classification, 3D model-based tracking and hypothesis assessment are described. The developed approach of 3D recognition of up to two vehicles requires a cycle time of 200 ms on a network of 7 transputers. The results achieved were demonstrated with synthetic and real images from motorway like scenes. (text in German language)