Cheap Joint Probabilistic Data Association Filters in an Interacting Multiple Model Design

This contribution presents an approach to fuse multiple sensors in an interacting multiple model design. Visual features like shadow and symmetry, treated as independent stand-alone virtual sensors, are employed for detection and tracking of vehicles for driver assistance tasks. Cheap joint probabilistic data association is utilised in order to account for the large amount of clutter present in the measurements provided by these sensors. Special attention is devoted to the different noise characteristics of the measurements. The individual sensors are considered in a sequential manner, leading to a versatile fusion architecture that allows easy integration of further sensor modules