基于街景图像的城市环境评价研究综述

城市环境评价研究传统上采用基于现场调研的方法,难以在大范围、精细化的尺度上进行评估。街景图像具有覆盖面广、能提供街道层级景观信息,且数据采集成本低的优势,为城市环境评价研究提供了大样本数据源和新的研究思路。人工智能技术的不断突破和其在各领域的应用,使得在大范围空间尺度上,基于街景图像进行城市环境评价研究成为可能。本文首先对城市环境评价常用的3种数据源（街景图像、遥感影像和地理标记社交媒体数据）进行对比分析,归纳街景图像在城市环境评价中的优势;然后,从方法学的角度把基于街景图像进行城市环境评价过程中使用的方法分为4大类别（基于图像分析的方法、基于统计分析的方法、基于人工智能的方法和基于空间分析的方法）;接着,从城市物理环境、社会环境、经济环境、美学环境,综述了街景图像在城市环境评价中的应用研究进展;最后,对现有研究成果进行了总结并对未来研究方向提出展望。

A Review of Urban Environmental Assessment based on Street View Images

Urban environmental assessment research has traditionally adopted a method based on field survey, which is difficult to evaluate on a large scale and refined scale. Street view image has a wide coverage, can provide street-level landscape and intuitively reflect the city facade information, and have the advantage of lower cost than on-site data collection, so it provides a large sample data source and new research ideas for urban environmental assessment. Different from the sky view of remote sensing image and the user interaction data of geo-tagged social media, street view image is more focused on recording stereoscopic sectional view of the city street level from the perspective of people, which can represent scenes seen or felt from the ground on a fine scale, so it is more suitable to replace on-site observation of urban environmental assessment. The continuous breakthrough of artificial intelligence technology and its application in various fields make it possible to conduct urban environmental assessment research based on street view image on a wide range of spatial scales. In this paper, we first described and compared three categories of data sources commonly used in urban environmental assessment including street view image, remote sensing image and geo-tagged social media data, and summarized the advantages of street view image in urban environmental assessment. Then we classified the methods used in urban environment assessment based on street view image into the following four categories : methods based on image analysis, statistical analysis, artificial intelligence and spatial analysis. Next, from the urban physical, social, economic and aesthetic environment, we summarized the research and application of street view image in urban environmental assessment. Finally, we pointed out the innovations, limitations and future research directions of the urban environmental assessment based on street view image. On one hand, the application of artificial intelligence represented by deep learning promotes the research progress of urban environmental assessment on large-scale and fine-scale. On the other hand, in the era of big data, the integration of data source represented by street view image, remote sensing image, and geo-tagged social media data will help promote urban environmental assessment research from multiple perspectives and multi-level.