Atualização de Documento

XIX SBSR

Formulário de atualização

O sistema de atualização, para eventuais correções, estará ativo até 30/11/2019.

Contato: bibdigital@inpe.br


Nome do Campo		Valor do Campo
Tipo de Referência		Artigo em Evento (Conference Proceedings)
Título do Trabalho	(*)(?)
	Não esqueça de clicar na opção "+" para informar os dados dos co-autores, se tiver.		v
1° Autor	(*)(?)
Grupo (1° Autor)
Afiliação (1° Autor)
e-Mail (1° Autor)
2° Autor	(*)(?)
Grupo (2° Autor)
Afiliação (2° Autor)
e-Mail (2° Autor)
3° Autor	(*)(?)
Grupo (3° Autor)
Afiliação (3° Autor)
e-Mail (3° Autor)
Páginas	(*)
Tipo Secundário
Palavras-Chave	(*)(?)
Resumo	(*)(?)	The present work is committed to conduct a comparative analysis between two supervised classifiers, Maximum Likelihood and Support Vector Machine, respectively parametric and non-parametric, for the classification of vegetal physiognomies using very high spatial resolution imagery, emphasizing the gain in performance with the accordingly increase in the number of attributes. The database consisted of pan-sharpened QuickBird images and transformed images derived from the original bands besides relief data obtained from the TOPODATA Project. The study area extends over a surface of 16 km2 and is located within the municipality of Nova Friburgo, in the mountainous region of Rio de Janeiro state. In total, four experiments were accomplished all of them combining the adopted classifier with a different number of attributes. In the first two experiments, only the four QuickBird spectral bands, previously subject to geometric and radiometric corrections, were used. In the remainder two experiments, eighteen input bands were employed. The Kappa indices obtained with the Maximum Likelihood classifier lied between 0.64 and 0.66, while those obtained for the Support Vector Machine ranged from 0.52 to 0.80. Considering the attained results, we concluded that the number of input bands does not meaningfully increase the accuracy of the Maximum Likelihood classifier, whereas this factor greatly influences the Support Vector Machine performance.
Tema	(*)
FileName	(?)
e-Mail (login)		simone