Distributed Data Fusion in a Ground Sensor Network

In order to fuse data from a distributed sensor system, different fusion architectures can be chosen. In this work, distributed data fusion has been applied to data from a ground sensor network. Measurements have been made using five seismic and five acoustic sensor nodes, each consisting of three sensors, on vehicles of five different types. The data were analyzed off-line to achieve tracking and classification of the vehicles. The tracking was made using standard Kalman filters, and classification was made by a combination of a Bayes’ classifier and voting. The information system, which made distributed data fusion possible, was based on a combination of intelligent agents and a distributed look-up table for data storage and retrieval. The results indicate that this system can operate on-line with a high accuracy in both tracking and classification.