Raster analysis of urban form

A new paradigm for investigating the environment consequences of urban texture is proposed. Using raster-based models and software algorithms derived from image processing the authors develop efficient methods of measuring geometric parameters and predicting radiation exchange. The possibilities of generating synthetic urban textures and of integrating cellular automata are explored. Results suggest the possibility of a raster-based urbanmodel to inform planning and design. DOI:10.1068/b2665 (1) A `raster model of urban form' may be defined as a two-dimensional, rectangular, pixel grid (such as a digital image) storing some urban attributes (such as building height). 2 Relation to previous work The analysis of raster maps with image-processing techniques is not new. It has been used in the geosciences for many years, with very successful results in the context of raster geographical information systems (GIS; for instance, see Burrough and McDonnell, 1998). Image-processing techniques are also regularly used in remote sensing, where they allow the extraction of features from collected digital imagery (see Jensen, 1986). Surprisingly enough, however, they have almost never been tested in the urban contextöprobably because of the difficulties that have existed so far in obtaining three-dimensional datasets in urban areas. A few examples can be found in our earlier work (Ratti, 2002, Richens, 1997). In this paper we build on that line of research and aim at presenting an introductory review of image-processing techniques applied to urban analysis for the planning community. 3 Image-processing basics for urban analysis The simplest possible representation of urban texture is a figure ^ ground map such as that shown in figure 1(a). If scanned, the resulting binary image tells us, for each pixel, whether the ground is built on or is open ground. If height information is included we obtain what geographers call a digital elevation model (DEM), which is an image where each pixel has a grey level proportional to the level of the urban surface. A DEM of an urban area in central London is shown in figure 1(b). This DEM was produced with a low-technology approach, based on scanning a digital ground map and colouring each pixel with a level of grey proportional to height (which was estimated from the number of storeys). Today, however, high-resolution remotely sensed DEMs are becoming increasingly available at low cost as a result of the rapid development of techniques such as laser altimetry or synthetic aperature radar (SAR). They can be commercially purchased for all major urban areas. Urban form represented as a DEM can be analysed with image-processing techniques using a simple package such as NIH Image.(2) Many of the ideas in NIH Image should be familiar from popular programs such as Adobe Photoshop. The difference is one of emphasis: Photoshop is concerned mainly with visual appearance, whereas NIH Image concentrates on measurement and analysis.