Crack Detection on 3D Surfaces Reconstructed from Drone Images of Wall Paintings

Abstract. Preserving wall paintings is a critical aspect of cultural heritage conservation, requiring precise methods for detecting and monitoring surface damage. This paper presents a methodology for crack detection on wall paintings using structure from motion to reconstruct 3D models from drone-acquired images. Highcurvature regions suggestive of cracks are identified through normal vector analysis and curvature computation, followed by binary dilation to improve the segmentation. The method highlights cracks effectively while ensuring accuracy and scalability. The results show a strong correlation between detected cracks and visible damage, showcasing the potential of this approach for non-invasive monitoring and conservation-restoration planning.