Quantifying the multidimensional urban perception of historic districts based on deep learning models

Abstract

Amid rapid urbanization, China faces the dual challenge of sustainable development and cultural preservation, with historic district revitalization emerging as a critical strategy. This study introduces a novel quantitative approach to evaluate urban vitality in historic districts by analyzing multi-dimensional urban perception data—including emotional dimensions such as aesthetic appeal, boredom, and depression—collected from street view imagery. Employing deep learning models enhanced with the Trueskill rating system, the research constructs a six-dimensional perception evaluation framework that accurately reflects the emotional experiences of both transient visitors and long-term residents. Focusing on Baihuazhou, a representative historic district, spatial analyses reveal a distinctive “high in the north and south, low in the east and west; high around the periphery, low in the center” vitality pattern, uncovering issues related to traffic congestion, insufficient public facilities, and suboptimal spatial functioning. Based on these findings, targeted recommendations are proposed in the realms of transportation re-planning, commercial layout diversification, and enhanced municipal management to improve overall urban functionality and resident well-being. Despite limitations such as a limited volunteer sample and incomplete spatiotemporal imagery data, this study offers a robust theoretical framework and methodological paradigm that contribute to both the deep understanding and pragmatic revitalization of historic urban districts within rapidly modernizing cities.