Low frequency feature extraction for target discrimination on a Biosonar dataset

PDF

Authors
  1. LeBlanc, T.
  2. Fawcett, J.
  3. Pailhas, Y.
  4. Williams, D.
Corporate Authors
Defence Research and Development Canada, Atlantic Research Centre, Halifax NS (CAN);Heriot-Watt Univ, Edinburgh (Scotland);Defence Science and Technology Lab, Porton Down (UK)
Abstract
Low frequency broadband sonar is an emerging technology in the field of underwater unexploded ordnance (UXO) detection and identification. These lower frequency acoustic signals can penetrate an object and allow one to extract target features based on both elastic effects and internal structure that will allow for more accurate target classification. One such sonar, Hydrason’s (UK) Biosonar is a broadband side-scan sonar that operates between 30 kHz and 130 kHz and uses dolphin inspired pulses. This paper examines data from trials conducted by the UK with Biosonar in the fall of 2012. Particular focus is given to the classification of a series of cylindrical targets with identical structures yet differing internal fillers. For these targets, numerous features are considered in order to differentiate between the targets i.e. their fillers. These features include time-domain features, spectral-based features, and aural-based features. Simple classification schemes are applied and the suitability of the features is discussed in the context of an ATR (automated target recognition) scheme.
Report Number
DRDC-RDDC-2014-N36 — External Literature
Date of publication
16 Mar 2015
Number of Pages
8
DSTKIM No
CA040059
CANDIS No
801095
Format(s):
Electronic Document(PDF)

Permanent link

Document 1 of 1

Date modified: