Radar and vision based data fusion - Advanced filtering techniques for a multi object vehicle tracking system

The robust and reliable detection of objects in the path of a vehicle is an important prerequisite for collision avoidance and collision mitigation systems. In this paper, an ego-motion compensated tracking approach is presented which combines radar observations with the results of a contour-based image processing algorithm. The approach is able to handle all uncertainties of the system in a unified way without analytical linearization by using the Unscented transform. By that, the covariances of the system can be estimated more accurately. The paper describes both the image processing and the state estimation algorithms. Furthermore, results of several practical tests are presented.