Population and dwelling unit estimates from space

Abstract

To increase the utility of satellite imagery as a source of cheap and current information for planning and managing cities some problems have to be resolved. One is information overload. As the spatial resolution of imagery increases to allow the identification of the individual components of urban form (street-blocks, buildings, gardens etc.), image interpretation (classifications into land-use categories or estimation of dwelling units) becomes less accurate. One answer is to adopt interpretation methods that use the increased information in a more detailed scene. This paper reports on attempts to measure the morphological patterns in an urban satellite scene and to use these for image interpretation. The interpretation task addressed is the estimation of residential dwelling units from the patterns discernible in high resolution satellite images of cities. The practical results include dwelling estimates that can be aggregated to any geographical unit of analysis, population estimates for cities and a dwelling density surface that can be categorised into any number of residential land-use classes.