JAILCD2026-005: Exploring the Relationship between the Built Environment of Waterfront Streets and Landscape Visual Aesthetic Quality Using Street View Images and Deep Learning

Abstract

Urban waterfront streets serve as a critical interface between natural environments and urban life, and their landscape visual aesthetic quality (LVAQ) plays an important role in shaping city image, spatial experience, and residents’ physical and psychological well-being. High-quality waterfront streetscapes can enhance urban attractiveness and promote tourism development. However, conventional approaches to visual aesthetic assessment are often constrained by high costs, time consumption, and limited spatial scalability. This study investigates the relationship between the built environment of waterfront streets and LVAQ using street-level imagery and deep learning techniques. Focusing on the Minato Mirai 21 district in Yokohama, street view images were employed in combination with artificial neural networks and semantic segmentation to quantify both LVAQ and street-level physical features. Correlation analysis was conducted to examine the associations between LVAQ and streetscape elements. The results indicate that physical features related to spatial openness (0.128), greenery (0.391), Vegetation diversity (0.220), landscape diversity (0.462) and walkability (0.182) exhibit more stable and significant statistical associations with LVAQ. By advancing the understanding of urban waterfront streets from a visual aesthetic perspective, this study provides a theoretical basis for improving waterfront streetscape quality and offers empirical support for human-centered urban planning and design.