Stručni rad
Relationship between vegetation indices and forest detection based on Landsat 5 images
Filip Sabo
; student, Fakultet tehničkih nauka, Univerzitet Novi Sad, Novi Sad, Srbija
Slobodan Pavlović
orcid.org/0000-0001-8415-4776
; student, Fakultet tehničkih nauka, Univerzitet Novi Sad, Novi Sad, Srbija
Dragana Popović
; student, Fakultet tehničkih nauka, Univerzitet Novi Sad, Novi Sad, Srbija
Sažetak
Remotely sensed data are widely used in order to determine changes on the Earth’s surface, disappearance or appearance of water areas, forests, agricultural areas, etc. This study aims to establish a relation between vegetation indices and forest detection based on multispectral satellite images with medium spatial resolution, i.e. how the combination of individual indices affects the extraction of forested areas and is that even possible. After establishing relations, it was possible to determine the deforested and forested areas.
The research was conducted using Landsat images with 30 meters spatial resolution and using the Erdas Imagine software. Individual
image bands (blue, red, green, near-infrared and mid-infrared) were used as input data. Rasters with certain pixel values, with which
forest classification was conducted, were generated using arithmetical combinations of image bands in the Model Maker. For each vegetation index, there are limit within which vegetation areas belong. The reason for using multiple vegetation indices are advantages and
disadvantages of individual indices. Some of them reduce background effects (bare soil) by using additional constants, while some have
less atmospheric impact on final results. Mainly, all of them use infrared bands because of high vegetation reflection of that spectrum.
Limitation for forest areas was established based on Ground Control Points (GCP) which represent true forest areas, and the classification
was performed using this limits. Lastly, vectorization (raster to vector conversion) of data was conducted in order to obtain areas. All of
this was applied during two summer periods, in 1986 and 2011.
Ključne riječi
Remote sensing; Landsat 5 Thematic Mapper; Vegetation indices; Erdas Imagine; Model Maker; Fruška Gora forests; Engineering classification
Hrčak ID:
125476
URI
Datum izdavanja:
26.6.2014.
Posjeta: 2.287 *