A computational framework for assessing the correspondence between 2D isovist-based and 3D perceptual visibility in exhibition spaces

Abstract

Plan-based isovist analysis has long provided a tractable means of quantifying spatial openness in architectural environments, yet it represents visibility as a horizontal section and fails to capture the volumetric, perspective-dependent qualities that govern what a visitor perceives at eye level. This divergence is particularly consequential in museum and exhibition spaces, where the visual relationship between the moving observer and spatially distributed artworks is central to both curatorial intent and visitor experience. Despite its practical relevance, a methodologically controlled comparison of plan-based and eye-level visibility within a shared scoring structure has not previously been undertaken in the architectural visibility literature for exhibition contexts. This thesis addresses this gap by proposing a dual-layered computational framework that evaluates two-dimensional isovist-based visibility alongside three-dimensional perceptual visibility derived from eye-level perspective projections. The study is conducted at the main exhibition gallery of CerModern Arts Center in Ankara, analyzing twelve vantage points established at ten-meter intervals along a visitor route, with observation height fixed at 1.75 m and view direction aligned with the direction of travel. The 2D model generates isovist polygons in Rhinoceros–Grasshopper, combining global spatial openness metrics with target-based indicators computed per artwork into a composite visibility score. The 3D model uses ID-pass rendering to extract the visible pixel area of each of the 87 modeled artworks from eye-level perspective images, computing screen-coverage ratios that capture the perspective-projected visual dominance of each target. Both models produce composite scores normalized to a common scale, and their absolute difference is categorized into three convergence classes: high convergence, partial convergence, and significant divergence. Results show that the two models converge where artworks are closely aligned with the travel direction and unobstructed by volumetric elements. Significant divergence emerges where high plan-level openness coincides with minimal perceptual content within the forward-facing visual cone, conditions under which the 2D model systematically overestimates visibility. The degree of divergence is structurally determined by the interplay of isovist area, occlusivity, and the three-dimensional positioning of exhibition content relative to the constrained visual cone, demonstrating that the 3D model acts as a corrective layer where plan-based analysis alone is insufficient. The framework makes three principal contributions to the field. First, it establishes a methodologically controlled basis for directly comparing plan-based and eye-level visibility within a shared scoring structure. Second, the 3D visibility layer introduces an adaptation of identity-pass (ID-Pass) rendering to architectural visibility analysis; by assigning unique color identities to each artwork and extracting visible pixel areas from eye-level perspective images, the method enables direct and objective measurement of artwork-level visual dominance in complex interior spaces, representing a novel methodological contribution to the field. Third, the framework provides practitioners with a diagnostic tool for identifying where plan analysis reliably approximates eye-level experience and where volumetric correction is needed, with direct applications in exhibition design and post-occupancy evaluation. Future research directions include user validation studies linking computed visibility scores to visitor recall and spatial learning outcomes, eye-tracking integration to replace the fixed-gaze model with behaviorally grounded gaze data, application of the framework to other building typologies such as educational facilities, hospitals, and public spaces, and systematic sensitivity analyses of the model parameters across varied spatial contexts.