Context Aware Multisensor Image Fusion for Military Sensor Networks using Multi Agent System

TL;DR

This paper introduces a context-aware multi-agent system for military sensor networks that improves multi-sensor image fusion by dynamically adapting to environmental and operational contexts, enhancing efficiency and accuracy.

Contribution

It presents a novel multi-agent framework that integrates context sensing, interpretation, and adaptive image fusion using wavelet transforms in military sensor networks.

Findings

Reduced fusion time and bandwidth usage.

Improved image fusion accuracy with lower mean square error.

Efficient resource utilization and energy consumption.

Abstract

This paper proposes a Context Aware Agent based Military Sensor Network (CAMSN) to form an improved infrastructure for multi-sensor image fusion. It considers contexts driven by a node and sink. The contexts such as general and critical object detection are node driven where as sensing time (such as day or night) is sink driven. The agencies used in the scheme are categorized as node and sink agency. Each agency employs a set of static and mobile agents to perform dedicated tasks. Node agency performs context sensing and context interpretation based on the sensed image and sensing time. Node agency comprises of node manager agent, context agent and node blackboard (NBB). Context agent gathers the context from the target and updates the NBB, Node manager agent interprets the context and passes the context information to sink node by using flooding mechanism. Sink agency mainly comprises of sink manager agent, fusing agent, and sink black board. A context at the sensor node triggers the fusion process at the sink. Based on the context, sink manager agent triggers the fusing agent. Fusing agent roams around the network, visits active sensor node, fuses the relevant images and sends the fused image to sink. The fusing agent uses wavelet transform for fusion. The scheme is simulated for testing its operation effectiveness in terms of fusion time, mean square error, throughput, dropping rate, bandwidth requirement, node battery usage and agent overhead.