Abstract

The high complexity of semantics extraction with automatic video analysis has forced the researchers to the generalization of mixed approaches based on perceptual and context data. These mixed approaches do not usually take in account the advantages and benefits of the data fusion discipline. This paper presents a context reasoning environment to deal with general and specific tracking problems. The cornerstone of the environment is a symbolic architecture based on the Joint Directors of Laboratories fusion model. This architecture may build a symbolic data representation from any source, check the data consistency, create new knowledge and refine it through inference obtaining a higher understanding level of the scene and providing feedback to autocorrect the tracking errors. An ontology-based prototype has been developed to carry out experimental tests. The prototype has been proved against tracking analysis occlusion problems.