Single-beam acoustic seabed classification systems are based, in general, on one of three methods: statistical segmentation based on acoustic similarity, measuring acoustic impedance, or inversion. Segmentation methods are now well established for seabed classification. They generate features that capture pertinent character and details of echoes, and then form groups that are acoustically similar, using clustering, neural networks, or genetic algorithms. This paper describes an acoustic segmentation seabed system for classifying underwater macro-algae by species. Echo features were generated from windows of the echo time series that were synchronized to the return from the seabed. Vegetation echoes precede the seabed echo while sediment interface and volume scattering follow it. In shallow surveys like these, the largest depths can be several times the least, so precise depth compensation is essential. Results from surveys in the Seto Inland Sea, Japan, are presented.
展开▼
