WIT Press


The Neural Method Of Sea Bottom Shape Modelling For The Spatial Maritime Information System

Price

Free (open access)

Paper DOI

10.2495/PORTS000221

Volume

51

Pages

9

Published

2000

Size

973 kb

Author(s)

A. Stateczny

Abstract

One of the basic problems of the spatial maritime information system is the three -dimensional numerical model of the sea bottom. It means depth interpolation of the sea bottom - in dependencies from geographical coordinates in any analysed point of an area. There exist a lot of classical methods of spatial modelling. In the article the neural method of marking depths in any investigated point of a reservoir and spatial modelling of the shape of the sea bottom is presented. Several types of neural networks were analysed and optimised to find a proper solution. Three of them, Multilayer Perception, Generalized Regression Neural Network (GRNN) and Radial Basis Function Network were chosen. The criterion of optimisation was accuracy of depth estimation. Trained neural networks can mark depth f

Keywords