Employing Artificial Intelligence Techniques in the Analysis of Morphological Characteristics of Heritage Elements in Iraq

Abstract

This research explores the application of artificial intelligence (AI) tools to the morphological analysis and digital documentation of heritage architectural elements, focusing on the muqarnas of the Abbasid Palace in Baghdad. A replicable workflow was developed that combines AI-based monocular depth estimation models, including MiDaS and Depth Anything, with point cloud generation, three-dimensional mesh reconstruction, and quantitative curvature analysis. High-resolution photographs were processed to produce detailed depth maps, which were transformed into dense point clouds and reconstructed into watertight meshes using the Poisson Surface Reconstruction algorithm. Per-vertex curvature metrics were extracted and visualized through color-coded maps and histograms, revealing a predominance of moderately low curvature regions interspersed with localized high-curvature zones corresponding to decorative recesses and stepped projections. Segmentation and classification of depth and curvature data enabled the surface to be divided into discrete morphological units, supporting a structured interpretation of its hierarchical design. These findings highlight the potential of integrating accessible AI-based tools with open-source 3D processing platforms to enhance the accuracy, objectivity, and reproducibility of heritage documentation. The methodology offers a scalable approach for analyzing similar historical elements and contributes to the broader discourse on digital preservation, conservation monitoring, and the interpretation of Islamic architectural heritage.