Semantic classification of human behaviors in video surveillance systems

The semantic analysis of the human behavior in video streaming is still an open issue for the computer vision research community, especially when real-time analysis of complex scenes is concerned. The researchers' community has achieved many progresses in this field. A popular class of approaches has been devised to enhance the quality of the semantic analysis by exploiting some background knowledge about scene and/or the human behavior, thus narrowing the huge variety of possible behavioral patterns by focusing on a specific narrow domain. Aim of this paper is to present an innovative method for semantic analysis of human behavior in video surveillance systems. Typically, this kind of systems are composed of a set of fixed cameras each one monitoring a fixed area. In the proposed methodology, the actions performed by the human beings are described by means of symbol strings. For each camera a grammar is defined to classify the strings of symbols describing the various behaviors. This system proposes a generative approach to human behavior description so it does not require a learning stage. Another advantage of this approach consists in the simplicity of the scene and motion descriptions so that the behavior analysis will have limited computational complexity due to the intrinsic nature both of the representations and the related operations used to manipulate them. This methodology has been used to implement a system to classify human behaviors in a scene. The results are discussed in this paper and they seem to be encouraging.