Identifying Socioeconomic Paterns in Urban Areas Trough Fuzzy Modeling and Object Oriented Image Classification

Population and incoming surveys provide a detailed description of urban areas revealing high quality information for planning and decision making. Traditional censuses are, however, very costly and are thus released only in very long time periods (average 10 years in Brazil). The integration of geo-referenced censitary data into Geographic Information Systems (GIS) allows the analysis of the spatial dependencies of socioeconomic data, but it is still dependent on the expensive surveys. Remote sensing imagery is a powerful tool for urban monitoring allowing the identification and detection of land use and landscape changes. High resolution images within modern classification algorithms allow the identification of the main urban features at a reasonable cost. This work presents a modeling approach to identify socioeconomic patterns in urban areas trough remote sensing images. The model relates image attributes to socioeconomic censitary data in a fuzzy rube based classification algorithm. The preliminary results of the proposed approach have been obtained using SPOT-5 images and income information provided by the Brazilian Geographic and Statistic Institute (IBGE). The fuzzy rule base is obtained by a recently proposed fuzzy rule based classifier induction algorithm [1]. The application under study is the monitoring of the development impact of the implantation of the Rio de Janeiro State Petrochemical Complex (COMPERJ), which was announced in 2006 and should start operation in 2010. This work presents preliminary results of a long-term project for approximately identify the urban changes within the COMPERJ area trough remote sensing images.