Segmentacija satelitskih snimaka šume za klasifikaciju tipa šume

master's thesis

• Title (english): Segmentation of satellite forest images for forest type classification
• Author: Sara Sičić
• Mentor: Marko Subašić (mentor)
• Committee member: Marko Subašić (predsjednik povjerenstva)
• Committee member: Tomislav Petković (član povjerenstva)
• Committee member: Juraj Petrović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-09-13, Croatia

Abstract

U ovom radu istražene su mogućnosti semantičke segmentacije slika šuma s ciljem klasifikacije tipova šuma na temelju označenih podataka pet različitih tipova šuma. Za problem segmentacije s ograničenom količinom podataka ispitane su dvije različite arhitekture: U-Net i VGG16, primijenjene u nekoliko različitih konfiguracija. Testirano je treniranje s augmentiranim podatcima, kao i uspješnost modela u odnosu na broj satelitskih snimaka korištenih kao ulaz u model te broj klasa koje model predviđa. Najuspješnijim se pokazao model VGG16. Daljnja poboljšanja mogla bi uključivati generiranje dodatnih podataka ili korištenje kvalitetnijeg skupa podataka.

Abstract (english)

This paper explores the potential of semantic segmentation of forest images for the classification of forest types based on labeled data for five different forest types. For the segmentation problem with a limited amount of data, two different architectures were examined: U-Net and VGG16, applied in several different configurations. The training with augmented data was tested, along with the model's performance in relation to the number of satellite images used as input to the model and the number of classes the model predicts. The VGG16 model proved to be the most successful. Further improvements could include generating additional data or using a higher quality dataset.

• Keywords: računalni vid; segmentacija slika; satelitske slike; klasifikacija; duboko učenje; neuronske mreže; VGG16; U-Net
• Keywords (english): computer vision; image segmentation; satellite pictures; classification; deep learning; neural networks; VGG16; U-Net
• Language: croatian
• URN:NBN: urn:nbn:hr:168:158788
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 22:05:45