Driving vision systems by communication

Abstract

Proposes a new multi-agent system architecture which is dedicated to building distributed computer vision systems. This architecture incorporates a communication language which provides a great degree of flexibility: firstly, the language permits the construction of flexible and self-organizing vision systems; secondly, it is capable of expressing complex facts and tasks; thirdly, it is simple, i.e. human-readable and writable; and lastly, the messages provided by the communication language can be interpreted efficiently. We describe in detail the basic concepts of the multi-agent system architecture, including the communication language as well as the agent architecture and the interaction strategy. As a testbed for the proposed architecture, we have modelled an object recognition system as a society of autonomous agents which organize themselves according to a given recognition task by employing communication.