Klasifikacija oblika galaksija primjenom algoritama u programskom jeziku Python

master's thesis

• Title (english): Galaxies morphology classification using algorithms in programming language Python
• Author: Filip Bošnjak
• Mentor: Goranka Bilalbegović (mentor)
• Committee member: Goranka Bilalbegović (predsjednik povjerenstva)
• Committee member: Vernesa Smolčić (član povjerenstva)
• Committee member: Katarina Jeličić (član povjerenstva)
• Committee member: Vibor Jelić (član povjerenstva)
• Committee member: Maja Planinić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Physics); Zagreb
• Defense date and country: 2022-03-04, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Physics

Abstract

Galaksije su gravitacijski povezani skupovi plina, kozmičke prašine, zvijezda i tamne tvari. Vjeruje se da većina velikih galaksija sadrži masivnu crnu rupu u središtu. U svemiru postoje milijuni galaksija od kojih je svaka posebna po svom obliku i veličini. Proučavanje galaksija, njihovog gibanja i transformacije je vrlo značajno za razumijevanje svemira kao cjeline. Donašanja teleskopska promatranja neba proizvode ogromnu i sve veću količinu podataka. Ručno označavanje morfoloških karakteristika galaksija je vremenski zahtjevan proces i podložan je pogreškama ljudske prirode. Astrofizičari i programeri pokušavaju riješiti ovaj problem automatizirajući proces klasifikacije i označavanja galaksija. U ovom radu se za analizu pripremljenih slika koriste konvolucijske neuronske mreže. To je najpopularniji algoritam strojnog učenja za analizu i obradu slika te detekciju objekata na njima. Zadatak je računalno razvrstavanje galaksija po obliku na temelju poznatih slika u vidljivom dijelu elektromagnetskog spektra. Ovakvim metodama strojnog učenja je moguće smanjiti pogreške u klasifikaciji galaksija. U metodičkom dijelu diplomskog rada je izložena priprema za održavanje nastavnog sata o algoritmima sortiranja.

Abstract (english)

Galaxies are gravitationally connected systems of gas, cosmic dust, stars and dark matter. It is believed that almost all big galaxies contain a massive black hole in their centers. There are millions of galaxies in the universe, and each of them is special in its shape and size. Study of galaxies, their motion and transformation plays an important role in understanding the universe as a whole. Current telescope observations of the sky produce a huge and growing amount of data. Manual labeling of morphological characteristics of galaxies is time consuming process, prone to human errors. Astrophysicists and programmers are trying to solve this problem by automatising the process of classification and labeling of galaxies. In this diploma thesis convolutional neural networks are used for the analysis of already collected images of galaxies. Convolutional neural networks algorithm is the most popular method in machine learning for the analysis and processing of images and objects detection on them. We describe the classification of shapes for previously known images of galaxies in the visible electromagnetic spectrum. With such methods of machine learning it is possible to reduce errors in the classification of galaxies. In the methodical section we present the preparation material for teaching sorting algorithms in high schools.

• Keywords: oblici galaksija; strojno učenje; konvolucijske neuronske mreže; nastava informatike za srednje škole: algoritmi
• Keywords (english): galaxies morphology; machine learning; convolutional neural networks; teaching computer science in high schools: algorithms
• Language: croatian
• URN:NBN: urn:nbn:hr:217:390583
• Study programme: Title: Physics and Computer Science; specializations in: Education; Course: Education; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: magistar/magistra edukacije fizike i informatike (magistar/magistra edukacije fizike i informatike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2022-03-21 08:54:51