Lageraiealade tuvastamine Sentinel-2 satelliitpiltide põhjal
Laen...
Kuupäev
2021
Kättesaadavus
08.09.2021
Autorid
Ajakirja pealkiri
Ajakirja ISSN
Köite pealkiri
Kirjastaja
Eesti Maaülikool
Abstrakt
Maailmas areneb tehnoloogia meeletu kiirusega. Sinna kuulub ka metsanduses ning
kaugseires kasutatav tehnoloogia, mida on võimalik ka omavahel siduda.
Antud töö eesmärk oli satelliitpiltide põhjal leida vahemikus 07.07.2018–21.06.2020
tekkinud lageraiealad Saaremaa riigimetsades, leida selle täpsus ning uurida vigade
tekitajaid. Selleks kasutati Sentinel-2A ja Sentinel-2B MultiSpectral Instrument poolt
tehtud pilte ning programme IDRISI Selva ja MapInfo Pro.
Töö tulemusena leiti riigimetsades asuvaid lageraiealasid kogupindalaga 657,74 ha.
Enamjaolt olid klassifitseerides leitud alade pindalad veidi suuremad kui eraldise tegelik
pindala, mis tähendab, et eraldiste servades olevate pikslite nurgad või servad ulatusid
eraldise piiridest välja. Selle vea peamiseks põhjustajaks olid heleda tooniga alad eraldiste
ääres, nagu näiteks metsateed või teised lagedad alad. Vastupidiselt oli ka eraldiste
servades tumedamaid alasid, näiteks varjud, mida klassifitseerimise käigus ei loetud
muutusena, mistõttu tuli eraldise pindala väiksem kui päriselt. Antud meetodit kasutades
sai lageraiealasid tuvastada keskmiselt 0,43 ha täpsusega. See oleneb siiski suuresti
eraldise pindalast, väiksemaid lageraiealasid on võimalik tuvastada täpsemalt ning
suuremate eraldiste tuvastamisel võib tekkida ka suurem viga.
In the world technology evolves with vast speed. This also includes the technology used in forestry and remote sensing, which can possibly be combined with each other. The purpose of this thesis was to find areas in the state forests in Saaremaa, where a clear cut had been carried out from 07.07.2018 to 21.06.2020, to ascertain its accuracy and examine the causes of the errors. For this, satellite photos taken by Sentinel-2A and Sentinel-2B MultiSpectral Instrument and programs IDRISI Selva and MapInfo Pro were used. As a result of the research, clear-cut areas in the state forests with a total area of 657,74 hectares were found. For the most part the regions found by classifying had a slightly bigger total area than the area of the actual forest allocation, which means that the corners or edges of some pixels on the edges of the allocations were out of the borders of said allocations .The main cause of this error were light toned areas at the edge of the allocations, for example forest roads or other clear areas. On the contrary, there were also darker toned areas at the edges of the allocations, for example shadows, that were not accounted as change while classifying, which is why the area of the allocation came out smaller than it actually is. By using the given method, clear-cut areas were possible to be determined with the average accuracy of 0,43 hectares. However, this largely depends on the area of the allocation, smaller clear-cut areas can be identified more accurately, and larger allocations can also be subject to a larger error.
In the world technology evolves with vast speed. This also includes the technology used in forestry and remote sensing, which can possibly be combined with each other. The purpose of this thesis was to find areas in the state forests in Saaremaa, where a clear cut had been carried out from 07.07.2018 to 21.06.2020, to ascertain its accuracy and examine the causes of the errors. For this, satellite photos taken by Sentinel-2A and Sentinel-2B MultiSpectral Instrument and programs IDRISI Selva and MapInfo Pro were used. As a result of the research, clear-cut areas in the state forests with a total area of 657,74 hectares were found. For the most part the regions found by classifying had a slightly bigger total area than the area of the actual forest allocation, which means that the corners or edges of some pixels on the edges of the allocations were out of the borders of said allocations .The main cause of this error were light toned areas at the edge of the allocations, for example forest roads or other clear areas. On the contrary, there were also darker toned areas at the edges of the allocations, for example shadows, that were not accounted as change while classifying, which is why the area of the allocation came out smaller than it actually is. By using the given method, clear-cut areas were possible to be determined with the average accuracy of 0,43 hectares. However, this largely depends on the area of the allocation, smaller clear-cut areas can be identified more accurately, and larger allocations can also be subject to a larger error.
Kirjeldus
Bakalaureusetöö
Metsanduse õppekaval
Märksõnad
bakalaureusetööd, klassifitseerimine, kaugseire, IDRISI, pildid, võrdlus, rasterandmed