AI Redrawing and Verification: Reliability Analysis Based on Image Generation and Qing Dynasty Official Embroidery Data

This study aims to address the challenges of structured representation and data annotation reliability concerning the regulatory knowledge of Qing Dynasty Rank Badges (Guanbu), in order to realize the digital reconstruction of complex cultural elements.We first constructed a Qing Dynasty regulatory knowledge network (167 nodes, 187 edges) and, based on this, proposed a “three-layer logical structure” (Rank \(\to \) Motif \(\to \) Detail). This structure confirms the hierarchical relationship in rank badge design and provides a formal foundation for the structured representation of regulatory knowledge. Regarding data reliability, Krippendorff’s α was employed to assess inter-annotator reliability. Results showed that the α value for military official badges reached 0.75–0.85, demonstrating high reliability. Conversely, the α value for civil official badges was lower (0.28–0.42), but their regulatory rules accounted for the highest proportion (63.9%), quantitatively revealing their higher regulatory complexity. The knowledge network and reliability analysis presented in this research provide a solid structural basis and data foundation for pattern generation models such as Stable Diffusion XL, effectively promoting the digital reconstruction and intelligent generation of cultural heritage elements.