Symbolic Translation of the Spatial Imagery of Fuzhou Ancient Houses and Its Driving Effect on Textile Industry Product Design

Architectural cultural heritage contains abundant geometric and chromatic information that can be systematically transformed into industrial design resources through computational modeling. This study proposes a semantic graph-based symbolic translation framework for converting the spatial imagery of Fuzhou ancient houses into textile product design elements. First, geometric contours are extracted using Canny edge detection and the Douglas–Peucker algorithm, while dominant color distributions are quantified in HSV space. A multi-layer semantic graph is then constructed from roof curvature, geometric proportions, symmetry characteristics, and color relationships. Node2Vec embedding and graph convolutional networks are subsequently employed to generate stable symbol vectors, which are further transformed into textile pattern matrices through frequency-domain texture synthesis based on fast Fourier transforms and parameterized repeating units. Experimental results demonstrate that the proposed framework achieves high symbolic stability and reliable color reproduction while maintaining satisfactory manufacturability for textile applications. Beyond regional cultural preservation, the combination of graph representation learning and frequency-domain pattern generation provides a computational paradigm for structured visual information encoding and periodic pattern synthesis. Such characteristics offer methodological references for digital design problems involving electromagnetic-inspired periodic structures, antenna surface pattern optimization, and intelligent engineering systems requiring spatial feature representation and frequency-domain analysis.