A Dual‐Symmetry‐Aware Method for Tilt Assessment on Chinese Ancient Pagodas: A Case Study on Yunyan Temple Pagoda

ABSTRACT Tilt assessment is a critical component in the health monitoring of cultural heritage structures. This study aimed to develop a dual‐symmetry‐aware method for accurate tilt assessment of Chinese ancient pagodas using multisource image fusion modeling. Traditional tilt measurement methods typically assess inclination by measuring coordinates of discrete feature points, but are subject to limitations including limited spatial representation and error accumulation. Moreover, the harsh external environments surrounding many structures further complicate tilt measurement. Oblique photogrammetry, which reconstructs 3D spatial structures through multiview imagery, enables rapid and cost‐effective documentation of geometric and textural information on building surfaces, and has thus been widely applied in the health monitoring of ancient structures in recent years. In this study, we propose a dual‐symmetry‐based (namely reflective and rotational symmetry) tilt assessment method for Chinese ancient pagodas using multisource image fusion modeling. Unlike existing methods that extract building central axes through thin point cloud slicing, we incorporate symmetry features common in Chinese pagoda architecture for 3D matching, thereby utilizing the 3D structural information in point cloud models more effectively. First, we propose an adaptive method to generate point cloud slices with progressively decreasing thickness. Second, a point cloud registration method was proposed which leverages a priori symmetry information from point cloud slices to perform 3D registration and symmetry parameter optimization, yielding more accurate centers for the slices. Finally, statistical analysis of density gradient variations for model points along the central axis enabled feature stratification of the architectural structure and refined tilt assessment. The proposed methodology was tested and validated on the Yunyan Temple Pagoda. Experimental results demonstrate that the proposed method generates more reliable center points for point cloud slices, thereby yielding more accurate central axis curves. Furthermore, the errors of our proposed method remain within 8.039% compared to manual tilt measurements, confirming its practical applicability and potential for deployment in broader scenarios.