基于街景图像的城市环境评价研究综述
作者简介:张丽英(1980-),女,博士生,主要从事时空数据挖掘和机器学习研究。E-mail:lyzhang1980@cup.edu.cn
收稿日期: 2018-07-04
要求修回日期: 2018-10-22
网络出版日期: 2019-01-20
基金资助
国家自然科学基金项目(41590845、41525004、41421001、41877523)
资源与环境信息系统国家重点实验室开放基金资助
A Review of Urban Environmental Assessment based on Street View Images
Received date: 2018-07-04
Request revised date: 2018-10-22
Online published: 2019-01-20
Supported by
National Natural Science Foundation of China, No.41590845, 41525004, 41421001,41877523
Supported by a grant from State Key Laboratory of Resources and Environmental Information System
Copyright
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。
张丽英 , 裴韬 , 陈宜金 , 宋辞 , 刘小茜 . 基于街景图像的城市环境评价研究综述[J]. 地球信息科学学报, 2019 , 21(1) : 46 -58 . DOI: 10.12082/dqxxkx.2019.180311
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.
Tab. 1 Street view image APIs表1 街景图像API |
街景API | 覆盖范围 | 图像最大分辨率(宽度高度) | 使用样例 |
---|---|---|---|
谷歌 | 114个国家及地区 | 20482048 | https://maps.googleapis.com/maps/api/streetview?size=400×400&location=40.720032,73.988354&fov=90&heading=235&pitch=10&key=YOUR_API_ KEY |
百度 | 中国372座城市 | 1024512 | http://api.map.baidu.com/panorama/v2?width=512&height=256&location=116.313393,40.04778&fov=180&ak=YOUR_API_KEY |
腾讯 | 中国296座城市 | 960640 | http://apis.map.qq.com/ws/streetview/v1/image?size=600480&pano=10011 022120723095812200&pitch=0&heading=0&key= YOUR_API_KEY |
Tab. 2 Comparison of street view image, remote sensing image and geo-tagged social media data表2 街景图像、遥感影像和地理标记社交媒体数据对比 |
数据类别 | 采样方式 | 优缺点 | 代表性研究 |
---|---|---|---|
街景图像 | 地面拍摄 | 优点:从微观和人的视角精细化记录城市街道层级的立体剖面景象;覆盖范围广、数据量大、成本低 缺点:数据在空间分布不均匀 | 社区环境[11]、城市安全感[21]、收入预测[22]、建筑特色[23] |
遥感影像 | 空中拍摄 | 优点:从宏观和高空鸟瞰的视角记录城市,覆盖范围广、数据在空间分布均匀 缺点:成本高、分辨率低 | 地表变化分析[12]、农作物识别[13]、空气质量评价[14]、灾情评价[15]、城市热环境变化[16] |
地理标记社交 媒体数据 | 网络用户发布 | 优点:用户交互的内容不仅有文本信息,还包含地理位置、时间、图像、视频、情感等信息;具有动态性、时效性和交互性; 缺点:数据稀疏、数据存在口语化、错误拼写和缩写、使用特殊符号等问题 | 空气质量[17]、台风灾害[18]、旅游景点评价[19]、城市风貌感知[20] |
Tab. 3 Urban environment evaluation based on street view image表3 基于街景图像的城市环境评价工作 |
类别 | 评价要素 | 数据集 | 方法 | 代表性研究 |
---|---|---|---|---|
物理环境 | 绿色植物、行人安全、行人道设施、机动交通、建筑、交通标志等 | 谷歌街景 | 相关性分析 泊松回归 机器学习 | 1.街景图像审计社区环境的可行性:Rundle[11],Badland[30],Clarke[45]; 2.城市安全性:Kronkvist[46],Li等[44],Mooney[47] 3.土地利用类型:Li[29] |
社会环境 | 汽车、人行道、行人、建筑、天空等 | 谷歌街景 | 机器学习 深度学习 | 1.人口分布与政治倾向:Gebru等[48];2.城市可步行性:Yin等[34],Hara等6],Yin[4];3.城市安全感:Porzi等[21],Li[27] |
经济环境 | 绿色植被、地面,建筑物、树、天空 | 谷歌街景 | 基于像元的图像分析机器学习 | 1.收入预测:Glaeser[22];2. 收入与物理环境的关系:Li[26],Arietta[49] |
美学环境 | 行道树、绿色植被、建筑物 | 谷歌街景百度街景 腾讯街景 | 机器学习 深度学习 图像分析 | 1.街道绿化:郝新华等[3],Berland[50],Li等[24]; 2. 城市风貌:Liu等[36],Cheng等[51] ,唐婧娴等[52]; 3. 建筑特色:Doersch等[53],Lee等[23] |
The authors have declared that no competing interests exist.
[1] |
[
|
[2] |
[
|
[3] |
[
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
Google Street View (GSV)[EB/OL].
|
[9] |
Baidu Street View (BSV)[EB/OL]. .
|
[10] |
Tencent Street View (TSV)[EB/OL]. .
|
[11] |
|
[12] |
|
[13] |
[
|
[14] |
|
[15] |
[
|
[16] |
[
|
[17] |
|
[18] |
[
|
[19] |
[
|
[20] |
[
|
[21] |
|
[22] |
|
[23] |
|
[24] |
|
[25] |
|
[26] |
|
[27] |
|
[28] |
|
[29] |
|
[30] |
|
[31] |
[
|
[32] |
[
|
[33] |
|
[34] |
|
[35] |
|
[36] |
|
[37] |
|
[38] |
|
[39] |
|
[40] |
|
[41] |
[
|
[42] |
[
|
[43] |
|
[44] |
|
[45] |
|
[46] |
|
[47] |
|
[48] |
|
[49] |
|
[50] |
|
[51] |
|
[52] |
[
|
[53] |
|
[54] |
|
[55] |
|
[56] |
|
[57] |
|
[58] |
|
[59] |
|
[60] |
|
[61] |
|
[62] |
|
[63] |
|
[64] |
[
|
[65] |
[
|
[66] |
|
[67] |
|
[68] |
|
[69] |
|
/
〈 |
|
〉 |