Over het archief
Het OWA, het open archief van het Waterbouwkundig Laboratorium heeft tot doel alle vrij toegankelijke onderzoeksresultaten van dit instituut in digitale vorm aan te bieden. Op die manier wil het de zichtbaarheid, verspreiding en gebruik van deze onderzoeksresultaten, alsook de wetenschappelijke communicatie maximaal bevorderen.
Dit archief wordt uitgebouwd en beheerd volgens de principes van de Open Access Movement, en het daaruit ontstane Open Archives Initiative.
Basisinformatie over ‘Open Access to scholarly information'.
| one publication added to basket [360071] |
Deep learning for internet of underwater things and ocean data analytics
Jahanbakht, M. (2022). Deep learning for internet of underwater things and ocean data analytics. PhD Thesis. James Cook University: Queensland. xvii, 203 pp.
|  |
| Beschikbaar in | Auteur |
|
Documenttype: Doctoraat/Thesis/Eindwerk
|
| Abstract |
The Internet of Underwater Things (IoUT) is an emerging technological ecosystem developed for connecting objects in maritime and underwater environments. IoUT technologies are empowered by a large number of deployed sensors and actuators. From scientific perspectives, these ubiquitous sensing devices are considered as data collecting tools, whichcan be augmented with machine intelligence and big data analytics both for automated monitoring and for prospective studies. In this thesis, we have comprehensively surveyed the IoUT and marine data analytics to address several research gaps. Using publicly available marine datasets, we have proposed three deep learning models for marine data timeseries. The proposed models include a timeseries forecasting ensemble of Deep Neural Networks (DNN) for sea surface temperature prediction, a next-frameprediction DNN framework for predicting total nitrogen in the Great Barrier Reef (GBR), and a Transformer-based next-frame prediction DNN framework for predicting total sediment in the GBR. Finally, an accurate and energy-efficient platform has been proposed for IoUT image processing for fish segmentation in realistic underwater video footages. This fast and low-bandwidth platform consists of a compressed DNN with low energyconsumption and real-time edge-based inferencing on an embedded GPU.The outcome of this thesis can facilitate developing tools for 1D and 2D spatiotemporal timeseries predictions. The proposed highly-accurate forecasting models can support decision makers to reach target water quality outcome in wide geographical areas like the GBR. Furthermore, the edge processing technique proposed in this thesis can take the marine video processing capabilities to the next level of intelligent in-situ IoUT systems. |
IMIS is ontwikkeld en wordt gehost door het VLIZ.
