Institutional Repository of Vilnius University
Daugiaspektrinių nuotraukų analizė naudojant giliojo mokymosi modelius /
Abstract
dc:descriptionDeep learning networks are often used in object recognition and segmentation tasks. These models are also applied when recognizing and measuring aquatic vegetation and algae blooms using multispectral images. The use of satellite imagery can help identify the vegetation and it’s concentration with higher accuracy than using standard RGB images. The main feature that sets this work apart is that any outside visual interferences are identified and removed from the edited satellite pictures in order to improve the quality of them. The purpose of this work is to perform the detection of aquatic plants and algae in satellite imagery and using deep neural networks remove and inpaint the clouds from them. The work examines the influence of Sentinel-2 satellite multispectral bands in detecting clouds and aquatic plants. Using an autoencoder and a GAN neural network the cloud mask generation model was fine-tuned in order to improve the results of cloud inpainting. The analysis of the obtained results are carried out using the initially created satellite image data set. The results of the study showed that the best results were achieved in the last training iteration, with a precision of 87% and recurcion of 86%.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Cinskas, Kostas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192846193&prefLang=en_US