A multi-agent architecture for intelligent distributed surveillance systems

The distributed surveillance systems include not only distributed array of cameras that offer wide area monitoring, but also a set of computer vision algorithms designed for scene analysis at multiple levels of abstraction. For building an intelligent surveillance system, the system requires the architecture for distributed computing environment to support real-time processing. This paper presents a multi-agent architecture for an intelligent distributed surveillance system. All processing modules in this system are encapsulated as agents. The proposed multi-agent architecture can be expressed by a multi-level concept. The surveillance system can be divided into 4 level tasks, which each level will be handled by one or several agents with different capabilities. Data management and transfer agent is a key module that supports real-time heterogeneous data stream sharing and exchanging among agents. Heterogeneous data and communication mechanism among agents are analyzed. The distributed surveillance system with the multi-agent architecture is currently applied as a test bed for the project that is based on meeting analysis and archive in intelligent meeting room in real-time.