Title
Klasifikace druhů dřevin s využitím hyperspektrálních dat získáných z bezpilotních leteckých prostředků
Code
SP2021/35
Summary
The traditional and oldest method of obtaining information on forest stands is field research. It is although very accurate, on the other hand, it is very time and money consuming. The advent of remote Earth exploration has revolutionized forest mapping and significantly reduced the time and cost required for data acquisition. The most appropriate way to monitor large continuous forest areas and identify trees is to use unmanned aerial vehicles. Commonly available commercial satellite images do not reach the required spatial resolution. Drones can be equipped with the necessary sensors, eg for taking high-resolution images in the infrared band (NIR), which is particularly sensitive to detect the health of vegetation, respectively (Dash et al.,2017). change in chlorophyll in the needles or leaves of the tree. With UAV, data sets can be obtained flexibly and quickly, with the high spatial and temporal resolution, if required.
The aim of this project is to use hyperspectral data from the UAV and to investigate the species composition and health status of the forest. The health assessment will be carried out in accordance with the methodology of Forest Health Assessment in the Czech Republic using satellite images (Lukeš, 2018). The following three methods will be used as classifiers, which were evaluated from the available literature as the most suitable (Dalponte et al., 2013): 1) support vector machine method 2) random forest 3) maximum likelihood method. The essence of the classification will lie in the division of the measured data file into certain classes according to a predefined classification rule. The classification rule can be determined on the basis of symptoms, which are significant and characteristic changes of the radiometric quantity depending on the change of the species or status parameter. There are four types of features, of which the most used are spectral features, and a distinction is made between spatial, temporal, and polarization features.
Literature:
M. Dalponte, H. O. Ørka, T. Gobakken, D. Gianelle and E. Næsset, "Tree Species Classification in Boreal Forests With Hyperspectral Data," in IEEE Transactions on Geoscience and Remote Sensing, vol. 51, no. 5, pp. 2632-2645, May 2013, doi: 10.1109/TGRS.2012.2216272.
Dash, J.P., Watt, M.S., Pearse, G.D., Heaphy, M., Dungey, H.S. Assessing very high resolution UAV imagery for monitoring forest health during a simulated disease outbreak, ISPRS Journal of Photogrammetry and Remote Sensing, Volume 131, 2017, Pages 1-14, ISSN 0924-2716, https://doi.org/10.1016/j.isprsjprs.2017.07.007.
Lukeš, P., Strejček, R., Křístek, Š., Mlčoušek, M. (2018) Hodnocení zdravotního stavu lesních porostů v České republice pomocí satelitních dat Sentinel-2. Ústav výzkumu globální změny AV ČR.
Start year
2021
End year
2021
Provider
Ministerstvo školství, mládeže a tělovýchovy
Category
SGS
Type
Specifický výzkum VŠB-TUO
Solver