Which pixel is a forest? Tree crown delineation using VHR images to estimate tree cover in landsat based classification

Determining the percentage of tree crown cover is extremely important to establish in advance which forest types can be classified with high resolution sensors such as Landsat. This paper describes the determination of a tree crown coverage threshold to define whether a pixel is classified as a forest or not. The methodology consists in the comparison of forest/non-forest classifications generated from Landsat images with tree crown cover maps obtained from PlanetScope very high resolution images, considering those pixels that exceed a given canopy cover threshold (eg. 5-10-15-...90-95-100%) as forest. The canopy coverage threshold was the one that minimized the difference between the Landsat classification and the maps generated from Planet images.

Bibliographic Details

Main Authors:	Banchero, Santiago, Veron, Santiago Ramón, De Abelleyra, Diego, Ferraina, Antonella, Propato, Tamara Sofia, Gomez Taffarel, Maria Cielo, Dieguez, Hernán
Format:	info:ar-repo/semantics/documento de conferencia biblioteca
Language:	eng
Published:	IEEE Geoscience and Remote Sensing Society 2021-07-12
Subjects:	Image Processing, Tratamiento de Imágenes, Vegetation Cover, Cubierta Vegetal, planet scope, alcance del planeta, Unsupervised learning, Aprendizaje sin supervisión,
Online Access:	http://hdl.handle.net/20.500.12123/10496 https://igarss2021.com/call_for_papers.php#
Tags:	Add Tag No Tags, Be the first to tag this record!