Detection and classification of marine targets buried in the sediment using structural acoustic features

Abstract

We present research on detection and classification of underwater targets buried in a saturated sediment using structural acoustic features. These efforts involve simulations using NRL’s STARS3D structural acoustics code and measurements in the NRL free-field and sediment pool facilities, off the coast of Duck, NC, and off the Coast of Panama City, FL. The measurements in the sediment pool demonstrated RVM classifiers trained using numerical data on two features—target strength correlation and elastic highlight image symmetry. Measurements off the coast of Duck were inconclusive owing to tropical storms resulting in a damaged projector. Extensive measurements were then carried out in 60 ft. of water in the Gulf using BOSS, an autonomous underwater vehicle with 40 receivers on its wings. The target field consisted of nine simulant-filled UXO and two false targets buried in the sediment and twenty proud targets. The AUV collected scattering data during north/south, east/west, and diagonal flights. We discuss the data analyzed so far from which we have extracted 3-D images and acoustic color constructs for 18 of the targets and demonstrated UXO/false target separation using a high dimensional acoustic color feature. Finally, we present related work involving targets buried in non-saturated elastic sediments. [This work is supported by ONR and SERDP.]