Underwater Unexploded Ordnance (UXO) Classification Using a Matched Subspace Classifier With Adaptive Dictionaries

Abstract

This paper is concerned with the development of a system for the discrimination of military munitions and unexploded ordnance (UXO) in shallow underwater environments. Acoustic color features corresponding to calibrated target strength as a function of frequency and look angle are generated from the raw sonar returns for munition characterization. A matched subspace classifier (MSC) is designed to discriminate between different classes of detected contacts based upon the spectral content of the sonar backscatter. The system is exclusively trained using model-generated sonar data and then tested using the measured Target and Reverberation Experiment 2013 (TREX13) data sets collected from a synthetic aperture sonar system in a relatively low-clutter environment. A new in situ supervised learning method is also developed to incrementally train the MSC using a limited number of labeled samples drawn from the TREX13 data sets. The classification results of the MSC are presented using standard performance metrics, such as receiver operating characteristic curve and confusion matrices.