An agent-oriented approach for image classification with ICA-color sift

This paper discusses the research conducted on developing a Multi-Agent System (MAS) for solving an image classification task. The aim of this research is to equip agents in MAS with reusable autonomous capabilities. The system provides a flexible framework for developing the communication aspects within an agent-oriented architecture to program agents that dynamically acquire functionality at runtime using event based messaging. In this research agents are equipped with unique image processing capabilities and required to interact and cooperate to achieve the goal. Complementary research on a variety of agent tools (specifically JACK, JADE and CIAgent) and communication languages (ACL, KQML, FIPA and SOAP) has been reviewed to glean knowledge that enables these agents to adapt those capabilities. The system has generated encouraging results.