Context Representation and Fusion via Likelihood Masks for Target Tracking

The inclusion of contextual information in low-level fusion processes is a promising research direction still scarcely explored. In this paper we propose a framework for integrating contextual knowledge in a fusion process in order to improve the estimation of a target's state. In particular, we will describe how contextual information can take the form of likelihood maps to be fused with the sensor's likelihood function in order to encode the dependence of the observation from the context.