May 10, 2011 | Urban Areas
Satellite based urban areas automatic classification
The goal of the experimental study is to implement a semi-automated unsupervised classification method to identify urban areas. Precise identification of urban areas is fundamental in order to obtain more precise estimates of people affected by an emergency as well as to carry out urban risk/vulnerability assessment analysis. Urban land is characterized by a heterogeneous spectral signature, due to the different surfaces it is composed by. Such characteristic preclude the use of traditional thematic classification algorithms for urban mapping, since they are generally based on assumptions on spectral homogeneity, that are very rarely met in urban environment. Preliminary results suggest to adopt a decision tree based approach where indexes, multitemporal analysis and spectral signature catalogues will be used as input data.