Human sequence evaluation:: towards knowledge-based scene interpretations

The term Video Surveillance refers to generic systems which describe the actions and interactions of different agents within complex scenes in real-time. In fact, video surveillance is a particular domain of the Image Sequence Evaluation framework, which provides a modular scheme to transform image data into high-level descriptions. In this paper, we are centered on human behavior description, denoted to as Human Sequence Evaluation. By considering the structure proposed in [9], we adapt the modules involved to cope with humans, and analyze the a-priori knowledge required to design such systems.