This study constructs a hierarchical knowledge graph for artistic-image annotation by bringing Chinese pictorial theories into structured dialogue with Western accounts of visual composition. Its central operation is the translation of qualitative art concepts—spatial organisation, dynamic relations, symbolism, pictorial language—into a systematic framework usable for dataset annotation and intelligent generation. The work responds to ambiguity and inconsistency not by stripping images down to generic labels but by expanding the conceptual resources available to describe them. Its cross-cultural ambition is methodologically significant: Chinese theories of space and symbolic construction are treated as generative components of the model rather than as decorative additions to a Western taxonomy. The graph thereby becomes an epistemic interface between visual knowledge and machine reasoning. Its broader bridge is to comparative aesthetics: computational legibility need not require cultural homogenisation, provided that the ontology preserves distinct conceptual lineages and makes their relations explicit rather than silently