基于深度学习的上海城市街景与景观美学公众认知研究

Study on the Public Perception of Shanghai Urban Streetscape and Landscape Aesthetics Based on Deep Learning

随着中国城镇化转型与城市更新的开展,服务城市大众美学认知与文化精神的城市风貌与景观塑造成为当前的迫切需求.以上海典型街景为例,结合问卷调查、深度学习与贝叶斯统计方法,构建具大范围评估应用潜力的公众审美感知模型,进行公众景观偏好研究.发现公众对于街景的不同场景审美感知差异较大;景观美学感知概念模型中关键景观要素为土地、建筑、树、墙;想象度、自然度、对比度和审美干扰度是影响公众对景观美学认知的重要指标.研究结果可进一步推广到大规模公众景观认知快速评估,为城市设计提供本土公众偏好美学理论的信息支持.

With the development of urban urbanization and urban renewal,urban style and landscape shaping serving the urban public's aesthetic cognition and cultural spirit has become an urgent demand.Taking examples of streetscapes in Shanghai,a combination of questionnaire survey,deep learning and Bayesian statistical methods are used to build a public aesthetic cognition model that has potential in evaluating the public landscape preference in a large scale.The findings suggest an obvious differences in the aesthetic evaluation of various streetscape scenes;the key landscape elements in the public aesthetic conceptual model are land,buildings,trees,and walls;imaginability,naturalness,contrast,and aesthetic interference are significant indicators that affect the public's perception of landscape aesthetics.The research results can be further applied to rapid assessment of large-scale public aesthetic preferences,providing support for the local public preference aesthetic theory for urban design.