A Methodology for Finding Uniform Regions in Spatial Data and its Application to Analyzing the Composition of Cities

Cities all around the world are in constant evolution due to numerous factors, such as fast urbanization and new ways of communication and transportation. However, the evolution of the composition of a city is difficult to follow and analyze. Since understanding the evolution of cities is the key to intelligent urbanization, there is a growing need to develop urban planning and analysis tools to guide the orderly development of cities, as well as to enhance their smooth and beneficial evolution. Urban patches which represent uniform areas of a city play a key role in studying the composition of a city, as different types of urban patches typically are associated with different functions, such as recreational areas and commercial areas. In order to analyze the changes of the composition of cities, a polygon-based spatial clustering and analysis framework for studying urban evolution is proposed in this thesis. A spatial clustering algorithm named CLEVER is used to identify urban patches that are clusters of polygons representing different elements of the city based on a domain expert's notion of uniformity, which has to be captured in a plug-in interestingness function. The analysis methodology uses polygons as models for spatial clusters and histogram-type distribution signatures to describe their characteristics. Finally, popular signatures are introduced that describe distribution characteristics, which occur frequently in contiguous sub-regions of a spatial dataset, and an approach is presented that identifies and annotates urban patches with popular signatures. Experiments on datasets of the city of Strasbourg, France serve as an example to highlight the usefulness of the methodology.