J. Łubczonek

Concept of Neural Model of the Sea Bottom Surface

This paper presents concept of neural model of the sea bottom surface, which could be implemented in marine numerical maps. The model consists of self-organizing network and set of trained networks used in surface approximation. Self-organizing networks firstly solves problem of division of data points into training subsets. Secondly it classifies any input data to one of approximating network. Set of approximating networks is used to calculate depth value at any coordinates of point within surface domain. An example illustrates reconstruction of surface based on neural model of bottom surface.

The accuracy of determination of radar shadows

Abstract Currently, the GIS software is often used to perform various tasks related to spatial planning. Exemplary use of GIS techniques is determination of marine radar shadows. Such analysis can be helpful during creating surveillance systems, which generally consist of several radar stations. But the use of standard analysis tools with the default functionality may not always yield satisfactory results. The reason for this, in the case of radar sensors, is the image distortion caused by the characteristics of the radar beam. This paper presents a modified method for the visibility analysis, which includes two basic radar image distortions - angular and radial distortion. Such improvement gives much more accurate results during determination of radar shadows then application of standard methods of viewshed analysis. Presented method was verified in the aspect of accuracy by using real radar image and radiolocation target. Further accuracy study took into consideration different DSM models used to perform viewshed analyses.