izpis_h1_title_alt

Rekonstrukcija površinske temperature morja z uporabo spektralne konvolucije
ID KLOPČIČ, MATIC (Author), ID Kristan, Matej (Mentor) More about this mentor... This link opens in a new window, ID Ličer, Matjaž (Co-mentor)

.pdfPDF - Presentation file, Download (1,66 MB)
MD5: B80099716D70567ACE395CC4C5F832D5

Abstract
Površinska temperatura morja je ključnega pomena za napovedovanje vre- mena, vendar pa oblaki zastirajo dele morja. Posledično je satelitska slika meritev nepopolna. Trenutno najboljša metoda za rekonstrukcijo SST je DINCAE2. Gre za samokodirnik s korakom izpopolnjevanja, ki z bili- nearno interpolacijo rekonstruira manjkajoče vrednosti. Razvili smo novo metodo za rekonstrukcijo manjkajočih meritev SST. Vmesne konvolucijske sloje v samokodirniku smo zamenjali z bloki hitre Fourierove konvolucije (FFC). FFC temelji na konvoluciji v frekvenčnem prostoru. Zastopa- nost vsake frekvence nosi informacijo s celotne slike, zato zaznavno polje pri konvoluciji v frekvenčnem prostoru že na prvem nivoju pokrije celotno sliko. Analizirali smo vpliv različnega števila blokov FFC, preskočnih povezav, ko- raka izpopolnjevanja, modula za združevanje značilk ter različnih cenilnih funkciji. Najboljša metoda je AEFFC, enojni samokodirnik z devetimi bloki FFC, brez preskočnih povezav. Njegova napaka na celotnem morju je za 0,5% večja od napake trenutno najboljšega modela DINCAE2, vendar je napaka rekonstruiranih delov pri naši metodi za 2% manjša.

Language:Slovenian
Keywords:hitra Fourierova konvolucija, samokodirnik, površinska temperatura morja
Work type:Bachelor thesis/paper
Typology:2.11 - Undergraduate Thesis
Organization:FRI - Faculty of Computer and Information Science
Year:2023
PID:20.500.12556/RUL-145239 This link opens in a new window
COBISS.SI-ID:149908483 This link opens in a new window
Publication date in RUL:13.04.2023
Views:436
Downloads:124
Metadata:XML RDF-CHPDL DC-XML DC-RDF
:
Copy citation
Share:Bookmark and Share

Secondary language

Language:English
Title:Reconstruction of sea surface temperature with fast Fourier convolution
Abstract:
Sea surface temperaure (SST) is crucial for accurate weather forecasting. However, clouds above the sea are preventing sensors on satellites to mea- sure SST beneath them. Therefore, some measurements are missing. Current state of the art method DINCAE2 consists of autoencoder with refine- ment step. It reconstructs the missing data with bilinear convolution. We proposed new method for SST reconstruction, which is based on fast Fourier convolution (FFC). In spectral space, every frequency covers the en- tire image. Consequently, using convolution in spectral space, the receptive field covers the entire image in the first layer of network. We have analysed the effect of different number of FFC blocks, skip connections, refinement step, feature fusion module and different loss functions. Our best method is AEFFC, which is an autoencoder with 9 FFC blocks without refinement step. State of the art method DINCAE2 has 0.5% lower error on the entire sea surface. Nevertheless, AEFFC has 2% lower error on the reconstructed surface.

Keywords:Fast Fourier convolution, autoencoder, sea surface temperature

Similar documents

Similar works from RUL:
Similar works from other Slovenian collections:

Back