Eddies in the ocean are a common and important occurrence that is essential to the movement of materials and energy in the marine environment. Therefore, the intelligent and exact identification of these eddies can substantially aid the advancement of our understanding of oceanography. The crowd is seeing a progressive improvement in the techniques used to identify these marine characteristics as a result of the continuous developments in cutting-edge deep learning technology. The Sea Surface Temperature (SST) data from the Copernicus Marine and Environment Monitoring Service (CMEMS) in the Atlantic Ocean are used in this study to present EddyNet, a unique deep-learning architecture developed for the automatic identification and categorization of ocean eddies. A pixel-wise classification layer is added to the convolutional encoder-decoder structure that serves as the foundation of EddyNet. The final product is a map with the same dimensions as the input, but each pixel is labeled as either "0" for non-eddy areas, "1" for anticyclonic eddies, or "2" for cyclonic eddies.
Le vortici nell’oceano sono un fenomeno comune ed importante che riveste un ruolo essen- ziale nel movimento di materiali ed energia nell’ambiente marino. Pertanto, l’identificazione intelligente ed esatta di questi vortici può notevolmente contribuire all’avanzamento della nostra comprensione dell’oceangrafia. Il pubblico sta assistendo a un progressivo miglio- ramento delle tecniche utilizzate per identificare queste caratteristiche marine grazie ai continui sviluppi nella tecnologia avanzata di apprendimento profondo. In questo studio, vengono utilizzati i dati sulla temperatura superficiale del mare (SST) forniti dal Servizio di Monitoraggio Marino e Ambientale Copernicus (CMEMS) nell’Oceano Atlantico per presentare EddyNet, un’architettura di apprendimento profondo unica sviluppata per l’identificazione automatica e la categorizzazione dei vortici oceanici. Uno strato di classi- ficazione pixel-wise viene aggiunto alla struttura encoder-decoder convoluzionale che funge da base di EddyNet. Il prodotto finale è una mappa delle stesse dimensioni dell’input, ma ogni pixel è etichettato come "0" per le aree non vorticose, "1" per i vortici anticiclonici o "2" per i vortici ciclonici.
Mesoscale eddy detection and classification from sea surface temperature maps with deep neural networks
SAFARI, MOHAMMAD MAHDI
2022/2023
Abstract
Eddies in the ocean are a common and important occurrence that is essential to the movement of materials and energy in the marine environment. Therefore, the intelligent and exact identification of these eddies can substantially aid the advancement of our understanding of oceanography. The crowd is seeing a progressive improvement in the techniques used to identify these marine characteristics as a result of the continuous developments in cutting-edge deep learning technology. The Sea Surface Temperature (SST) data from the Copernicus Marine and Environment Monitoring Service (CMEMS) in the Atlantic Ocean are used in this study to present EddyNet, a unique deep-learning architecture developed for the automatic identification and categorization of ocean eddies. A pixel-wise classification layer is added to the convolutional encoder-decoder structure that serves as the foundation of EddyNet. The final product is a map with the same dimensions as the input, but each pixel is labeled as either "0" for non-eddy areas, "1" for anticyclonic eddies, or "2" for cyclonic eddies.File | Dimensione | Formato | |
---|---|---|---|
Master_Thesis_modified.pdf
accessibile in internet per tutti
Dimensione
4.15 MB
Formato
Adobe PDF
|
4.15 MB | Adobe PDF | Visualizza/Apri |
Master_Thesis_modified_without_images.pdf
accessibile in internet per tutti
Dimensione
557.8 kB
Formato
Adobe PDF
|
557.8 kB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/209798