Introducing a Perceptual–Spatial Landscape Planning Model (PSLPM) for cultural landscapes’ route optimization

Abstract

As cultural landscapes increasingly evolve within urban environments, a persistent gap has emerged between their heritage spatial characteristics and public heritage value perception, undermining urban identity and sustainability. This study introduces the Perceptual–Spatial Landscape Planning Model, an integrative framework that positions route optimization as a strategic planning intervention to bridge this gap. Taking Chengde Mountain Resort as a case, we constructed a spatial network of 144 scenes and collected 43,879 user-generated social media comments to quantify public perception. We achieved perceptual–spatial coupling through: 1) a few-shot learning paradigm based on a large language model generates five distinct perception scores of each scene; 2) the resulting scores are then integrated with heritage spatial characteristics in graph neural networks using graph attention networks and deep graph infomax. Based on the coupling process, the optimized routes were generated by a multi-objective evolutionary algorithm. The optimized routes outperformed official and greedy baselines, achieving higher spatial diversity, perceptual coherence, lower variance, and broader coverage of marginal yet meaningful scenes. These routes avoided overconcentration in perceptual hotspots as conventional tourism planning while enhancing interpretive richness across various spaces. This model provides strong contextual transferability, offering interpretable framework for integrating public perception into cultural landscape planning, thereby advancing both the methodological rigor and spatial intelligence of cultural landscapes’ development in urban environments.