Prepoznavanje japonskega dresnika iz posnetkov multispektralnega fotoaparata

magistrsko delo št.: 103/II. GIG

Patricija Vrhovšek (Author), Dejan Grigillo (Mentor), Klemen Kozmus Trajkovski (Co-mentor)

Abstract

Japonski dresnik je tuja invazivna rastlinska vrsta, ki zaradi velike razširjenosti in hitrega razmnoževanja izpodriva avtohtone rastline, zavira njihovo rast ter tako zmanjšuje biodiverziteto. Pri reševanju te problematike je ključno zaznavanje dresnika v začetni fazi rasti in manjših sestojev, ter spremljanje dinamike njegovega širjenja in oblikovanje napovednih modelov prostorske razširitve. V magistrski nalogi je predstavljena uporaba objektno usmerjene nadzorovane klasifikacije za zaznavanje japonskega dresnika iz ortofotov, izdelanih iz posnetkov multispektralnega fotoaparata MicaSense RedEdge-M, nameščenega na daljinsko vodeni letalnik. Izbrano območje ob vodotoku Mali graben, kjer se nahajajo večji in manjši sestoji japonskega dresnika, smo posneli v treh različnih fenoloških dobah japonskega dresnika in z dveh različnih višin. Za klasifikacijo te invazivne rastlinske vrste na visokoločljivih multispektralnih slikovnih virih smo uporabili metodo objektne klasifikacije na podlagi učnih vzorcev in na podlagi pravil, pri kateri japonski dresnik od domorodnih rastlin ločimo z uporabo spektralnih, teksturnih in prostorskih atributov.

Keywords

geodezija;magistrska dela;GIG;multispektralni fotoaparat;japonski dresnik;objektna klasifikacija;daljinsko vodeni letalnik;

Data

Language:	Slovenian
Year of publishing:	2020
Typology:	2.09 - Master's Thesis
Organization:	UL FGG - Faculty of Civil and Geodetic Engineering
Publisher:	[P. Vrhovšek]
UDC:	528.7/.8:630(497.4)(043.3)
COBISS:	32350723
Views:	631
Downloads:	198
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Detection of Japanese knotweed from multispectral images
Secondary abstract:	Japanese knotweed is a foreign invasive plant species that is, due to its increased prevalence and rapid propagation, displacing native plants and inhibiting their growth, thus reducing biodiversity. Some of the key factors in solving this problem, its detection in the early growth and small plant formation phase, the control of its spread dynamics and the development of a spread prediction model. The master's thesis examines the use of object-based image classification for the detection of Japanese knotweed from orthophotos, produced from images taken by the MicaSense RedEdge-M multispectral camera, which was installed on an unmanned aerial vehicle. The selected area by the Mali Graben watercourse, where bigger and smaller plant formations of Japanese knotweed grow, was recorded in three different phenological phases of the plant's development and in two different heights. For the classification of this invasive plant species on high-resolution multispectral images we used the example-based and rulebased object-oriented image analysis, where Japanese knotweed is distinguished from native plants with the use of spectral, texture and spatial attributes.
Secondary keywords:	geodesy;master thesis;multispectral camera;Japanese knotweed;object-based classification;unmanned aerial vehicle;
Type (COBISS):	Master's thesis/paper
Study programme:	0
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Fak. za gradbeništvo in geodezijo
Pages:	X, 40 str.
ID:	11890303