Please use this identifier to cite or link to this item: http://dx.doi.org/10.18419/opus-10323
Authors: Mattes, Dennis Frederic
Title: Analysis of waveforms in the satellite altimetry by using neural networks
Issue Date: 2019
metadata.ubs.publikation.typ: Abschlussarbeit (Master)
metadata.ubs.publikation.seiten: XIII, 99, XXVI
URI: http://elib.uni-stuttgart.de/handle/11682/10340
http://nbn-resolving.de/urn:nbn:de:bsz:93-opus-ds-103407
http://dx.doi.org/10.18419/opus-10323
Abstract: The in situ data of inland water bodies is only limited and is declining lately. At the same time, it is more and more important to monitor the inland water bodies, since the climate is changing rapidly. To handle this problem, space born sensors are used more and more. One of the possibilities is to use satellite altimetry, which was previously designed for measurements over the oceans. Thereby, the satellite is transmitting a radar signal towards the earth surface at nadir. This signal is reflected by the ground back to the satellite. By doing so, it estimates the surface height with the runtime of the signal. However, caused by the fast changing terrain over the inland, more noise is included and lead to errors in the height estimation. To solve this, retracker are applied which analyse the received signal and estimate the correct runtime. In this thesis, a new approach will be presented which aims to use neural networks for the retracking purpose. The advantage is that neural networks can learn the characteristic pattern of the signals and then find this pattern during the retracking process. Thereby two approaches are developed, one which uses solely a neural network and a second one, which uses the results of the neural network as an input for an algorithm. They are then applied to different study areas to analyse their performance. It could be shown that the neural networks can estimate the water height well so that a reasonable water height time series can be created. Thereby, the neural network approach shows better results than the algorithm. At the end also the transferability of the neural networks could be shown. Thus, one can use a trained neural network also on other water bodies as which are used for training.
Appears in Collections:06 Fakultät Luft- und Raumfahrttechnik und Geodäsie

Files in This Item:
File Description SizeFormat 
MasterArbeit_DennisMattes.pdf13,22 MBAdobe PDFView/Open


Items in OPUS are protected by copyright, with all rights reserved, unless otherwise indicated.