| Autor: |
Abedi, Roya, Bonyad, Amir Eslam, Moridani, Alireza Yousefi, Shahbahrami, Asadollah |
| Předmět: |
|
| Zdroj: |
International Journal of Image & Data Fusion; Dec2018, Vol. 9 Issue 4, p287-301, 15p |
| Abstrakt: |
k Nearest neighbour (kNN) is a non-parametric method to integrate satellite image and ground inventory data in order to estimate accurate information and becoming increasingly popular in natural resources and forest sciences. The main objective of this study was to evaluate the IRS and Landsat 8 satellite images data for forest attributes (volume, basal area and tree density) estimation by kNN method in the north forest of Iran. Our results suggested k ≤ 10 and Euclidean and Fuzzy distance metrics were the best for our study area. We concluded that Landsat 8 image improved the quality of predictions by reducing the error in kNN algorithm predictions. Also, the greatest amount of root mean square error estimated for tree density and the least was for basal area and volume, respectively in all images. Therefore, kNN method showed the best efficiency for basal area but have smaller efficiency for tree density and volume. We found that main original spectral bands have the best results versus normalised difference vegetation index and fusion data. The strength of kNN was the possibility to predict several variables consist of continuous and categorical variables. Therefore, we suggested studying forest cover types as a categorical variable by this method in diverse natural hardwood species in the north forest of Iran. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
