Information Fusion in Signal and Image Processing: Major Probabilistic and Non-Probabilistic Numerical Approaches

Chapter 10

Multi-Agent Methods: An Example of an Architecture and its Application for the Detection, Recognition and Identification of Targets

The purpose of vision systems is to allow an understanding of the observed scene, based on diverse data obtained from images. This scene is often characterized by a non-homogenous, highly variable environment and any observation condition. Obviously, this general nature leads to very complex systems, including the need for scalability, for the ability to provide intermediate results, the integration of uncertain knowledge, or the possibility to adapt through the choice of strategies, operators and parameters.

The problem of vision that we have just introduced involves the detection, recognition and identification (DRI) of targets such as military ground vehicles, boats and aircraft. In this operational context, we have to be capable of making a decision as quickly as possible, in order to assess the threat. A DRI system needs to constantly search for the relevant information among a wealth of useless elements of information. We propose a method based on multi-agent concepts to solve this problem. At each instant, a population of agents works in parallel in the environment. Each agent has descriptive and operational knowledge at its disposal to allow it to elaborate its own strategy and to conduct processes. The set of results obtained is stored in a world model shared by all the agents.

This chapter is organized as follows: in section 10.1, ...