Abstract
Detection of unexploded ordnance is challenging in the underwater environment, particularly when object burial occurs. A capability to detect buried targets has been demonstrated previously using TNO’s MUD low frequency sediment-penetrating sonar and other similar sonars. However, the high clutter rates encountered in practice have the potential to impose severe operational limitations in absence of a robust capability to distinguish targets from clutter. To this end, we are taking the initial steps towards development of an automatic target recognition algorithm for detecting targets and suppressing clutter in low-frequency sub-sediment sonar imagery. The initial implementation presented in this paper uses a previously developed wavelet shrinkage algorithm to suppress the background reverberation, followed by automatic thresholding and segmentation to isolate individual seafloor objects for subsequent extraction of their acoustic signatures. We show preliminary detection results from the MUD-2011 data set.