Analyzing architectural space: identifying salient regions by computing 3D isovists

Spatio-cognitive properties - including the capacity to evoke a sense of place or support wayfinding – are amongst the most important considerations in the design of large urban and architectural spaces. Both wayfinding and spatial identity rely on the capacity of a space to be noticeably distinct; a property that is called saliency. However, there are relatively few experimental techniques to identify locations that are rich in such visual properties and those that do exist were typically developed in the 1980s using only two-dimensional viewshed geometry. In contrast, this paper describes the use of 3D isovists to develop a salient region estimation technique for architectural and urban environments. The underlying method of saliency estimation is based on principal component analysis that is used to compare geometric properties of the areas surrounding a vantage point. Statistical summaries of the 3D isovists are compared in the principal component space to differentiate the monotonous regions from the ones that are more visually distinct. The experimental results reported in this paper are developed using a model of the Villa Savoye to demonstrate that 3D isovists can be used to determine the extent to which an environment supports a capacity for wayfinding. The paper makes two major contributions to architectural computational analysis; first, it demonstrates a consistent, stable 3D isovist method and second, it proposes a quantitative technique for detecting regions with strong visual characteristics.