Topološka analiza vremenskih slik

MUBI, GREGOR (author), Mramor Kosta, Nežka (mentor)

V nalogi je predstavljena analiza vremenskih podatkov s pomočjo računske topologije. Cilj naloge je analizirati vremenske slike in razbrati, ali lahko izpostavimo nekatere parametre, ki bi izboljšali klasifikacijo vlažnih gmot. V ta namen smo skozi zaporedje slik zasledovali, kako se rojevajo in združujejo vlažne gmote nad področjem na sliki. Uporabili smo topolosko metodo, ki se imenuje vztrajnost. Za analizo podatkov smo pridobili vremenske radarske slike. Za računanje in vizualizacijo podatkov smo napisali približno 20 skript v delovnem okolju MATLAB. Slike smo predstavili kot kubične komplekse, kjer so sivinske vrednosti podane v ogliščih kompleksa. Eden od rezultatov dela je tudi identifikacija parametrov, ki temelji na topoloških metodah in bi bil lahko eden od kazalnikov za napovedovanje toče in drugih izjemnih vremenskih pojavov.

• Language: Slovenian
• Keywords: računska topologija, vztrajna homologija, kubični kompleksi, vztrajni diagrami, vremenski podatki, napovedovanje vremena
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2015
• PID: 20.500.12556/RUL-72083
• Publication date in RUL: 24.08.2015

Secondary language

• Language: English
• Title: Topological analysis of weather images

Abstract

In this thesis computational topology is used to analyse weather data. The goal is to analyse weather images and discern, if it is possible to expose certain parameters, which could improve humid mass classification. For this reason we were observing how those masses appear and merge on sequences of images. We used topological method named persistence. For data analysis we obtained weather radar images. We also wrote around 20 scripts in computing environment MATLAB, which served for data computation and visualization. Images are presented as cubical complexes, where grayscale values are in vertexes of the complexes. One of the results is also an identification of parameter, which is based on topological methods and could be used in hail forecast and other exceptional weather forecasts.

• Keywords: computational topology, persistent homology, cubical complexes, persistent diagrams, weather data, weather prediction