Building-level change detection from large-scale historical vector data by using direct and a three-tier post-classification comparison

Abstract

Historical change information at the building level in urban areas is crucial for policy and resource management, especially in countries with densely population and quick building construction. In this paper we present a multi-level building change detection framework using large-scale historical vector and address-based data. This approach is fundamentally different to the traditionally ones which purely use remotely sensed images and are often limited in identifying functional characteristics. Ordnance Survey’s (OS) MasterMap in the UK has been taken as an example of the large-scale vector data. The buildings features are extracted for two years and are compared to identify modified, demolished, and un-changed ones. To quantify buildings’ functional changes, an earlier developed classification methodology was used by extracting cartomteric and spatial properties of buildings and linking contextual information from address-based data. The case study in Manchester, UK shows that the proposed approach can successfully identify building changes at multiple levels. The change detection framework presented here closely addresses how to use large-scale and existing data sources to create a historical land use database. Moreover, this framework is computationally robust and is applicable to other areas without losing its integrity within and outside the UK, where large-scale structured data sets are available.