Multi-agent based framework for target tracking using a real time vision system

Abstract

This paper presents a multi-agent based framework for target tracking. We exploit the agent-oriented software paradigm with its characteristics that provide intelligent autonomous behavior together with a real time computer vision system to achieve high performance real time target tracking. The framework consists of four layers; interface, strategic, management, and operation layers. Interface layer receives from the user the tracking parameters such as the number and type of trackers and targets and type of the tracking environment, and then delivers these parameters to the subsequent layers. Strategic (decision making) layer is provided with a knowledge base of target tracking methodologies that are previously implemented by researchers in diverse target tracking applications and are proven successful. And by inference in the knowledge base using the user input a tracking methodology is chosen. Management layer is responsible for pursuing and controlling the tracking methodology execution. Operation layer represents the phases in the tracking methodology and is responsible for communicating with the real-time computer vision system to execute the algorithms in the phases. The framework is presented with a case study to show its ability to tackle the target tracking problem and its flexibility to solve the problem with different tracking parameters. This paper describes the ability of the agent-based framework to deploy any real-time vision system that fits in solving the target tracking problem. It is a step towards a complete open standard, real-time, agent-based framework for target tracking.