Integration of contextual information for tracking refinement

The exploitation of contextual information can bring several advantages to fusion systems at different levels. Although very promising, this topic is still a scarcely explored. In particular, the inclusion of contextual information in low-level fusion processes has not received much attention in the literature. In this paper we propose a framework for integrating contextual knowledge in a multisensor fusion process in order to improve the estimation of a target's state for tracking. Context will be here encoded in the form of likelihood maps to be fused with the sensors' likelihood functions. The framework presented here allows employing either discounting or pruning strategies for assessing the reliability of sensor observations. In our preliminary experiments, the inclusion of context has provided better accuracy in simulated multisensor tracking scenarios.