Object-based image analysis using QuickBird satellite images and GIS data, case study Belo Horizonte (Brazil)

Abstract

Brazil, like many other developing countries is demanding modern IT tools that would allow to monitor, to analyze and to intervene in the city planning process. Urban areas, especially metropolitan regions like Belo Horizonte (Minas Gerais State Capital, Brazil) are highly complex and its’ diagnosis is quite difficult. With the advent of high-resolution satellite systems, such as Quick-Bird, object-based image classification techniques became efficient procedures for the analysis and mapping of land use/land cover. This new image classification paradigm uses context relations, hierarchy and fuzzy logic. A multi-temporal analysis, considering also ancillary data, allows spatial inferences generating information to subsidize urban planning. In this study, spatial inferences were made for two quarters of Belo Horizonte using QuickBird OR-Standard scenes from 2002 and 2004, working with object-based image classification techniques and considering geological, geotechnical and urban legislation data on a GIS. At both scenes ortho-rectification was done using a rigorous model. Following information were generated considering the growth of two quarters (Belvedere and Buritis) in the timeframe considered: a land use/land cover map, detection of irregular land occupation, of areas with risks for slope slipping/erosion as well as an analysis of potential damage to population and property.