A graph-based algorithm to define urban topology from unstructured geospatial data

Abstract

Interpretation and analysis of urban topology are particularly challenging tasks given the complex spatial pattern of the urban elements, and hence their automation is especially needed. In terms of the urban scene meaning, the starting point in this study is unstructured geospatial data, i.e. no prior knowledge of the geospatial entities is assumed. Translating these data into more meaningful homogeneous regions can be achieved by detecting geographic features within the initial random collection of geospatial objects, and then by grouping them according to their spatial arrangement. The techniques applied to achieve this are those of graph theory applied to urban topology analysis within GIS environment. This article focuses primarily on the implementation and algorithmic design of a methodology to define and make urban topology explicit. Conceptually, such procedure analyses and interprets geospatial object arrangements in terms of the extension of the standard notion of the topological relation of adjacency to that of containment: the so-called ‘containment-first search’. LiDAR data were used as an example scenario for development and test purposes.