Finisterra - Revista Portuguesa de Geografia (Jun 2007)
Classificação de uso do solo urbano através da análise linear de mistura espectral com imagens de satélite
Abstract
URBAN LAND USE CLASSIFICATION THROUGH SATELLITE IMAGES SPECTRAL MIXTURE LINEAR ANALYSIS. Urban environments are heterogeneous by nature. Hence, to allow quantitative studies it’s necessary to simplify them in combinations of basic land use/cover materials. The Ridd’s VIS model (1995) is a conceptual representation that allows simplifying urban environments through the combination of three basic components: vegetation (V), impervious surface (I) and soil (S). The majority of urban uses can be interpreted as a combination of those three basic components.The VIS analysis allows to disclose that the mainstream of urban features has it’s own VIS signature, which is difficult to detect through pixel-by-pixel based classifiers. This work examines the land use/cover characteristics of the Great Lisbon Metropolitan Area (GAML) using sub-pixel classification techniques, mainly linear spectral unmixing (LSU), developing a conceptual model to characterize the occupation standards. The LSU ability to measure the physical composition of urban morphology is also explored and tested. In this paper we use Landsat 7 ETM+ multispectral images and SPOT 5 HRVIR, evaluating at the same time the land use/cover signatures evolution and the effect of spatial resolution differences on the same signatures measurements.
