Repository hosted by TU Delft Library

Home · Contact · About · Disclaimer ·

Long range echo classification for minehunting sonars

Publication files not online:

Author: Theije, P.A.M. de · Groen, J. · Sabel, J.C.
Publisher: Nexus Media, Ltd
Place: Swanley, Kent
Institution: TNO Defensie en Veiligheid
Source:Proceeding of the European conference on Undersea Defence Technology, UDT Europe 2006, June, 27-29, Hamburg, Germany
Identifier: 223238
Keywords: Defence


This paper focesus on single-ping classification of sea mines, at a range of about 400 m, and combining a hull mounted sonar (HMS) and a propelled variable-depth sonar (PDVS). The deleoped classifier is trained and tested on a set of simulated realistic echoes of mines and non-mines. As the mines ate too far away for image-based classification, only the information in the time series and their frequency spectra is used, representing a set of features. The classifier is based on finding the best combination of features via so-called Fischer Linear Mapping. The results show that reliable classification performance can be achieved, nearly 90% correct classification at 20% false classification, especialy after combining the features measured by the HMS and PDVS