Izdelava in analiza časovne vrste satelitskih posnetkov različnih virov : magistrsko delo

Hegediš, Valentin (author), Oštir, Krištof (mentor)

Z združevanjem podatkov iz različnih satelitskih sistemov imamo možnost zagotovitve večje časovne gostote podatkov. Pri tem se pojavi vprašanje, kako posnetke med seboj združiti, če so med posnetki različnih sistemov razhajanja. V magistrskem delu smo združevali dva vira visokoločljivostnih satelitskih posnetkov – Sentinel-2 in PlanetScope. Iz omenjenih posnetkov smo izdelovali rastre NDVI in jih med seboj harmonizirali s parametri regresijskih premic njihovih grafov raztrosa. Te parametre smo uporabili na rastrih NDVI PlanetScope in jih tako približali rastrom NDVI Sentinel-2. Po harmonizaciji so se vrednosti rastrov NDVI PlanetScope in Sentinel-2 bolje ujemale, kar smo v magistrski nalogi prikazali tako na samih rastrih NDVI, njihovih histogramih kot tudi na časovnih vrstah NDVI vrednosti. Pri vrednostih parametrov je oblačnost igrala veliko vlogo, zato smo rezultate prikazali za različne stopnje jasnosti neba. Z interpolacijo parametrov harmonizacije smo nad posameznimi območji rabe GERK izdelali časovne vrste spreminjanja vrednosti NDVI. Izvedbo korakov v magistrski nalogi smo avtomatizirali s programskima jezikoma Python in PL/pgSQL.

• Language: Slovenian
• Keywords: satelitski posnetki, harmonizacija, časovne vrste, Python, PostgreSQL
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FGG - Faculty of Civil and Geodetic Engineering
• Publisher: V. Hegediš]
• Year: 2019
• PID: 20.500.12556/RUL-113291
• UDC: 528.837:629.783(043.3)
• COBISS.SI-ID: 9017441
• Publication date in RUL: 19.12.2019

Secondary language

• Language: English
• Title: Time Series Analysis of Satellite Images from Different Sources : master thesis

Abstract

Combining the images obtained from different satellite systems we have the possibility of obtaining data points with greater temporal resolution. This presents us with a challange of combining satellite images that are somewhat different in their values. In our master thesis we describes the process of combining two sources of high-resolution satellite images – Sentinel-2 and PlanetScope. We produced NDVI rasters and harmonized them using linear regression trend line parameters obtained from scatter plots of PlanetScope and Sentinel-2 NDVI rasters. With those parameters we altered the values of PlanetScope NDVI rasters to make them more in line with values of Sentinel-2 NDVI rasters. We showed that there was a noticeable improvement between values of un-harmonized input and harmonized output PlanetScope data in comparison to the NDVI values of the referenced Sentinel-2 data. We used the parameters we obtained in the harmonization process to make combined time series from PlanetScope and Sentinel-2 data on different areas of land use that we obtained from GERK. We automated the whole process using Python and PL/pgSQL.

• Keywords: satellite images, harmonization, time series, NDVI, Python, PostgreSQL