Classificação em área urbana apoiada em imagens aéreas e dados LIDAR

Abstract

In urban centers, land use mappings are widely used for land planning and management. For this, several laws were created to indicate guidelines for land use and occupation to guarantee the quality of life to citizens and the resources maintenance for future generations. In this context, developing a methodology for a better understanding of the urban landscape is an useful tool for government decisions. Therefore, the proposal of this research is to elaborate an object-oriented classification based on fuzzy modeling that still considers three-dimensional elements to better categorize the ground cover types. In addition a validation step was developed to analyze the thematic quality of the proposed methodology. The study area for this research is the Planalto neighborhood, in the city of Belo Horizonte, because its variability of coverage and land use. Topics about the extraction of information from high resolution images, the LIDAR, the object-oriented classification, the fuzzy logic and the data validation based on stratified random samples are the guidelines for this research. With this, a detailed segmentation was elaborated for analysis scale; class descriptors where weighted and a classification methodology considering three-dimensional attributes from lidar and cartographic databases were consolidated. A coverage and use map was created, considering the possibility of application in other locations. Among the overall thematic validation indexes, the kappa coefficient stands out, and the value found was 62%, a result considered substantial. Other indexes by class were analyzed to evaluate the quality of the classification. In comparison to the official classification used in Belo Horizonte city, the research presented better results, mainly in relation to the mapping detail level, the validation methodology and the land cover and use classes, that allows to observe urban phenomena such as verticalization, soil waterproofing, etc.