In this paper, we propose a framework for the assessment, interpretation and understanding of situations in a visual surveillance scenario. An upper level processing layer exploits abductive reasoning over data provided by low level detection, classification and tracking algorithms. The system discussed is Description Logics driven and benefits from the computability of first-order logic semantics together with the manageability characteristics of ontology based systems. Event occurrence frequency is taken into account to focus on “anomalous facts,” by combining a priori knowledge, provided by domain experts, with statistical information incrementally gathered through the assessment of the environment under surveillance. The framework is aimed at supplying security system operators with a set of the most probable explanations of observed facts to improve and speed up the decision process.
