Visual Narrative of Multi-Ethnic Pattern Symbols from the Perspective of the Chinese Nation’s Community Consciousness: Intertextuality Between Southwest Ethnic Minority Brocade and Dunhuang Caisson Ceiling

Abstract

With the increasing application of electromagnetic sensing, wireless imaging systems, and intelligent information transmission technologies in digital cultural heritage preservation, efficient cross-media analysis of visual symbols has become essential for reliable semantic understanding and knowledge dissemination. To address the insufficient analysis of cross-temporal and cross-spatial intertextuality between southwestern ethnic minority brocades and Dunhuang caisson ceilings, as well as the lack of quantitative symbol-sharing mechanisms and narrative reconstruction methods, this paper proposes a dual-channel feature extraction and association modeling framework. Mask R-CNN is first employed to extract pattern contours and generate cultural semantic annotations. A dual-branch neural network is then designed to quantify the color-texture characteristics of brocades and the geometric structures of caisson ceilings, while maximum mean discrepancy is adopted to align heterogeneous visual features in a unified latent space. Furthermore, cosine similarity and symbol co-occurrence analysis are combined to construct an intertextual association graph, and an LSTMbased sequence decoder reconstructs visual narrative paths to reveal spatiotemporal cultural transmission patterns. Experimental results show that the symbol sharing rate between Miao brocade and Dunhuang grid patterns reaches 0.92, while the cross-media sharing rate of sacro-symbolic motifs is 0.52, demonstrating significant intertextual relationships and cultural convergence. The proposed framework provides an effective computational paradigm for multi-ethnic visual narrative analysis and offers methodological support for intelligent information processing and electromagnetic-enabled digital heritage applications.