Selection of SOM parameters for the needs of clusterization of data obtained by interferometric methods

Abstract

The article presents a detailed analysis of parameter settings of self-organizing map (SOM) for the clusterization of bathymetric data obtained using interferometric techniques. Clusterization using SOM is one of the stages of a new a geodata reduction method being currently researched by the authors for the purpose of a bathymetric map construction. In the research authors used data obtained by GeoSwath+.-interferometric sonar system. Test data gathered from the area of 100m² included 3760 data points. During the tests the authors focused primarily on setting individual network parameters in the course of network training and also on their importance in the clustering of bathymetric data. A total of forty-eight different scenarios of SOM parameter settings were tested. In the article detailed analysis of the obtained results is presented with an emphasis on the use of SOM in future studies related to new geodata reduction method.

Precise Bathymetry as a Step Towards Producing Bathymetric Electronic Navigational Charts for Comparative (Terrain Reference) Navigation

“…For example, for the purpose of navigation safety, users need to visualise the shallowest depths in true positions with a spacing dependent on the display scale. Reduction of bathymetric geodata with a focus on minimum depth has been described in previous articles (Wlodarczyk-Sielicka and Stateczny, 2015; Wlodarczyk-Sielicka et al, 2016).…”

MSIS Image Postioning in Port Areas with the Aid of Comparative Navigation Methods

“…The model of seabed was created using highdensity bathymetric data obtained from the interferometric system GeoSwathPlus for the area of interest. The issues of processing of high-density data obtained using remote sensing methods are discussed in [30][31][32][33][34][35][36][37][38][39][40][41], among other sources.…”

The Use of an Artificial Neural Network to Process Hydrographic Big Data during Surface Modeling

“…After starting the competition, the node l with the closest distance is selected as the winner. In the next stage, the weights w of the winner is updated using the learning principle (24). The weight vector w for the lth node in the sth step of training is denoted as w s l and the input vector for the ith training simple is denoted as X i .…”

“…In the next step, we focused on two parameters: the distance between each neuron and its neighbors, and initial neighborhood size. In this case, we fully based our parameters on previous studies that were published in [24]. The distance was measured using different functions: Euclidean distance, link distance, box distance, and Manhattan distance.…”