Significant lines: measuring and representing architecture for computational analysis

Whenever a computational or mathematical analysis of architectural space or form is undertaken, a range of assumptions must be made about which information in the building, survey drawing or CAD model to include. For example, shape grammar analyses of architectural forms are typically focussed on the macro-scale geometry of plans while tectonic details (like column and wall junctions) are ignored. Similarly, space syntax analysis will often disregard formal or ornamental modelling in favour of producing a simplified plan before the computational analysis takes place. Fractal analysis and semantic programming techniques for architecture also rely on drawings or CAD representations of buildings, from which selective data is derived for review and mathematical processing. But in many of these various cases, there are no clear conventions describing which lines, in a plan, model or drawing, are significant for architectural analysis. This paper commences with a critical overview of the range of different ways buildings are represented, prior to computational analysis. The logic inherent in such representational choices is then discussed and a basis for decision making on architectural data is proposed. To support this discussion, the paper uses the computational method of fractal analysis to examine several architectural elevations, in each case demonstrating how different architectural information produces different mathematical results. Finally the paper offers a heuristic or model for determining which lines in specific architectural images are significant and should thereby be subjected to analysis of this type