Autor(es): Coelho, Ana Margarida ; Sousa, Adélia ; Gonçalves, Ana Cristina
Data: 2023
Identificador Persistente: http://hdl.handle.net/10174/34159
Origem: Repositório Científico da Universidade de Évora
Assunto(s): Multiresolution Segmentation; WorldView-3; Forest Species; Object-Oriented Classification
In Portugal forest occupies 35,8% of the national territory, 23% corresponding to conifers and 64% to broadleaved species. According to the national forest inventory of 2015, the main forest species are eucaliptus (Eucalyptus spp.) representing 26%, cork oak (Quercus suber) and maritime pine (Pinus pinaster) with 22% of the forest area. The soil land use/cover monitoring has a pivotal role in the natural resource’s management, in the climatic changes study, in the territory and forest planning and in the sustainable development. Currently, the wide range of data sets derived from remote sensing allows the cartography of forest land use at different spatial scales, contributing to the evaluation of the protection and conservation of the forest areas, and the quantification of biomass and carbon. This study objective is the delimitation and identification of the crowns per forest species in Alentejo region, based on satellite images of high spatial resolution 0,50 m (Worldview-3), using the multiresolution segmentation and object-oriented classification. In the multiresolution segmentation several set of input variables were tested (original bands, bands, vegetation indices, principal components), with different threshold parameters values (colour, form and dimension) in order to isolate the objects in the image. After, object-oriented classification was used to isolate the tree crowns from the other classes.
