Visual Knowledge Representation of Moving Scenes

Abstract

A framework for the representation of visual knowledge, with particular attention to the analysis and the representation of scenes with moving persons and objects is described. Our aim is to integrate in a principled way the models developed within the artificial vision community, and the propositional knowledge representation systems developed within symbolic AI. We claim that this integration requires the introduction of a missing link between these two classes of representations. In our model, such a role is played by the notion of conceptual space (CS), a representation in which information is characterised in terms of a metric space. We illustrate our proposal by referring to an experimental setup, based on a vision system that takes in input real images of moving people.