A machine learning approach for estimation of shallow water depths from optical satellite images and sonar measurements

Z. Vojinovic*, Y. A. Abebe, R. Ranasinghe, A. Vacher, P. Martens, D. J. Mandl, S. W. Frye, E. Van Ettinger, R. De Zeeuw

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

29 Citations (Scopus)

Abstract

There has been a rapid growth in the field of remote sensing and its various applications in the area of water management. Nowadays, there are several remote sensing techniques that can be used as a source to derive bathymetry data along coastal areas. The key techniques are: sonar (sound navigating and ranging), LiDAR (light detection and ranging) and high-resolution satellite images. The present paper describes a method which was developed and used to create a shallow water bathymetry data along the Dutch side of Sint Maarten Island by combining sonar measurements and satellite images in a nonlinear machine learning technique. The purpose of this work is to develop a bathymetry dataset that can be used to set up physically-based models for coastal flood modelling work. The nonlinear machine learning technique used in the work is a support vector machine (SVM) model. The sonar data were used as an output whereas image data were used as an input into the SVM model. The results were analysed for three depth ranges and the findings are promising. It remains to further verify the capacity of the new method on a dataset with higher resolution satellite imagery.

Original languageEnglish
Pages (from-to)1408-1424
Number of pages17
JournalJournal of Hydroinformatics
Volume15
Issue number4
DOIs
Publication statusPublished - 2013

Keywords

  • Bathymetry
  • Machine learning
  • Remote sensing
  • Satellite images
  • Sonar measurements
  • Support vector machine

Fingerprint

Dive into the research topics of 'A machine learning approach for estimation of shallow water depths from optical satellite images and sonar measurements'. Together they form a unique fingerprint.

Cite this