Traditional architectural decorative patterns are increasingly reused in contemporary design, yet the link between object selection and design generation often remains experience-driven: public perceptual differences are rarely formalized, and evaluation outcomes seldom constrain generative decisions. This study proposes a perceptual demand-driven layered filtering and design response model (PD–LFDR) that treats traditional architectural decorative patterns as comparable and traceable design resources. Perceptual inputs from multiple stakeholders are converged via Kansei-based semantic aggregation into four core dimensions—symbolism, heritage authenticity, recognition and regionality—and are organized as a perceptual evaluation matrix. Grey relational analysis (GRA) is then applied using an expected perceptual level as the reference sequence to identify representative pattern samples suitable for design intervention. An empirical study on decorative patterns from Shaanxi vernacular dwellings demonstrates a closed-loop workflow: (i) first-round GRA filters representative theme samples, (ii) a second-round GRA selects operable minimal gene units, and, under a unified parametric rule set and a traceable two-layer parameter basis (parameter domain definition and parameter selection), (iii) multiple alternatives are generated and re-evaluated through a third-round GRA to support scheme selection. Robustness checks indicate stable rankings under moderate parameter and weight variation, improving interpretability, reproducibility, and decision efficiency for the computational translation of regional cultural visual resources.
Wang et al. (Fri,) studied this question.