IDC 99: Abstract

Detection and Classification For Unattended Ground Sensors

Abstract:

This paper describes the application of statistical estimation and multisensor fusion techniques to the detection, classification and localisation of targets with unattended ground sensors (UGS). Data have been collected for personnel and vehicles moving near a seismic sensor, and processed to allow two features to be obtained. An EM algorithm is used to estimate parameters of a Gaussian mixture model representing the signal data, resulting in improved detection and classification compared to the sensor's built-in algorithm. A design is given for multisensor fusion of a set of UGS sensors, in which multiple detections by a single sensor are used to estimate the closest distance of the target to the sensor, based on an assumption about the target speed, and these estimates are associated to a set of predefined trajectories which the target is assumed to follow. A multiple hypothesis scheme is used to update the probabilities of the trajectories as the target moves past the sensors through the region under surveillance.

BibTeX Reference:

@InProceedings{Goodman[1999],
	author =       {Graham L Goodman},
	title =        {Detection and Classification for Unattended Ground Sensors},
	booktitle =    {Proceedings of Information Decision and Control 99},
	pages =        {419--424},
	year =         {1999},
	editor =       {Robin Evans and Lang White and Daniel McMichael and Len Sciacca},
	address =      {Adelaide, Australia},
	month =        {February},
	publisher =    {Institute of Electrical and Electronic Engineers, Inc.},
	url =          {http://www.eleceng.adelaide.edu.au/ieee/idc99/abstracts/goodman2.html}
}