Open Journal Systems

INTEGRAÇÃO DE IMAGENS DE SATÉLITE E DADOS COMPLEMENTARES PARA A DELIMINATAÇÃO DE UNIDADES DE PAISAGEM USANDO UMA ABORDAGEM BASEADA EM REGIÕES

SELMA REGINA ARANHA RIBEIRO, HANS PETER BÄHR, JORGE SILVA CENTENO

Abstract



No artigo são apresentados os resultados de um estudo que tem como objetivo
delimitar unidades de paisagem, para fins de mapeamento de solos, utilizando dados
digitais advindos de sensoriamento remoto e informações auxiliares de agentes que
podem atuar nos processos de formação dos solos, como o relevo e a rede de
drenagem. A metodologia utilizada baseia-se na análise de regiões com
características similares para a separação das unidades de paisagem, sendo a
similaridade definida em termos espectrais e das variáveis topográficas
consideradas. O esquema de classificação utilizado analisa não apenas um pixel
isoladamente, mas dividie a imagem em segmentos, considerados como objetos
primitivos da imagem. Tais objetos são posteriormente analisados como unidades,
sendo possível levar-se em conta atributos espaciais derivados das regiões.
A principal vantagem desta metodologia encontra-se associada ao fato de que o uso
de regiões como unidade de análise é mais apropriado para a descrição de unidades
de paisagem do que uma abordagem pixel a pixel e ao mesmo tempo, a inclusão de
informações complementares é um fator determinante na separação de classes muito
similares em termos espectrais.

Integration of satellite imagery and ancillary data for landscape units using a
region based approach

Abstract


This article presents the results of a study that aims at mapping landscape units, for
soil mapping purposes, using remote sensing digital data and ancillary information
derived from other agents that can influence the processes of soil formation, as the
relief and the drainage net. The approach is based on the analysis of regions with
similar characteristics, being the similarity defined in terms of spectral information
and other topographical variables. Therefore, the image has to be divided in small
segments, considered primitive objects of the image. A classification is later used,
that analyses a whole region, instead of an isolated pixel, and decides the most
suitable class for the region based on the information contained in its set of pixels.
The main advantage of this approach is associated with the fact that the use of
regions as unit instead of pixels is more appropriate for the description of the
landscape and that the inclusion of complementary information can contribute to
the discrimination of spectral very similar classes.