Seabed Classification Using Artificial Neural Networks and Other Nonparametric Methods

Four nonparametric statistical classi cation methods are evaluated for the problem of segmentation and} classi cation of the seabed based upon data recorded from a multibeam echo sounder., Two of the methods are the multilayered perception (MLP) and radial basis function (RBF) neural networks.The other two methods are'polynomial regression mod els (PRM) and a B spline modelling technique (ASMOD).The classi cation is based upon a set of features extracted from the strength of the echo recorded within a rectangular region (pixel) of the seabed.Feature extraction is_not covered in this article, but we refer to another article on this subject in these proceedings.Test results for two different geographical regions are presented, and the performance of the different classi ers is discussed.