Detekcija objekata na SAR snimkama niske rezolucije metodama dubokog učenja

master's thesis

• Title (english): Object detection in low-resolution SAR images using deep learning methods
• Author: Dino Tognon
• Mentor: Marko Bosiljevac (mentor)
• Committee member: Marko Bosiljevac (predsjednik povjerenstva)
• Committee member: Sonja Grgić (član povjerenstva)
• Committee member: Dario Bojanjac (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2023-07-11, Croatia

Abstract

Rad započinje kratkim uvodom u radarsku tehnologiju te je objašnjena motivacija za izradu sustava za automatsku detekciju objekata na SAR snimkama. Ukratko je opisan princip rada neuronskih mreža, konvolucijskih neuronskih mreža te DBSCAN algoritma grupiranja. U zadnjem poglavlju detaljno je opisan razvijeni sustav za detekciju objekata na radarskim snimkama niske rezolucije. U sklopu opisa sustava, opisan je razvijeni algoritam za lokalizaciju objekata, objašnjen je način izgradnje i odabira klasifikatora za korištenje u sustavu te su prikazani rezultati izvođenja algoritma i navedene sve izmjerene performanse na skupovima podataka različitih apertura. Na samom kraju rada dane su smjernice za daljnji rad i unaprjeđenje sustava.

Abstract (english)

The thesis begins with a short introduction to radar technology and the reasoning behind creating a system for automatic detection of objects in SAR images. The second chapter briefly describes the working principles of neural networks, convolutional neural networks and the DBSCAN grouping algorithm. The final chapter describes the developed system for detecting objects on low-resolution radar images in detail. As part of the system description, the developed algorithm for object localization and the method of building and selecting a classifier for use in the system are described and explained. Also, the results of the algorithm execution are presented and performance metrics on different aperture data sets are listed. At the very end, a few guidelines are given for further improvement of the system.

• Keywords: SAR; CNN; detekcija objekata; klasifikacija objekata; DBSCAN
• Keywords (english): SAR; CNN; object detection; object classification; DBSCAN
• Language: croatian
• URN:NBN: urn:nbn:hr:168:442656
• 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: 2023-10-11 10:46:35