Automatic detection and analysis of internal waves on SAR images

Abstract

Internal waves features, like all mesoscale oceanographic features, are an important aspect of the ocean circulation. They are responsible for an important energy transfer mechanism, and have many implications in oceanographic engineering developments. Currently the extraction of internal waves information from satellite images is usually done from the human interpretation of the grey tone pattern visible in the images, which is a subjective, labour-intensive and time consuming task.

In this research a new method for the automatic detection of internal waves’ signatures present in SAR images has been developed. The automatic detection technique uses two different approaches. One is based on wavelet transform and statistical texture descriptors. The classifications have been implemented using principal component analysis, the K-Nearest Neighbour technique and the multi-layer perception. The second approach is based on shape discrimination. The geometry, orientation and position of the different edges found within the image are used to distinguish the presence of internal waves. Along with a reduction in man power and analysis time, this new technique offers the means to analyse internal waves. Data sets of internal waves based on a number of criteria can easily be created. The users can then use the information to study the internal wave’s dynamics or the internal wave conditions in a given place, which could be of value for offshore development.

The results of the research outlined in this thesis have demonstrated that the combination of either textural analysis with classifier or edge geometry analysis can provide the recognition and a primary analysis of internal wave signatures. This technique would therefore provide an appropriate starting point for the development of an operational recognition tool.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata