| 2,766 | 53 | 570 |
| 下载次数 | 被引频次 | 阅读次数 |
随着当前城市规划与设计的精细化转型，人本尺度下的街道空间品质特征研究日益受到广泛关注。作为影响街道空间品质的关键要素之一，街道渗透率即街道底层门窗洞口面积占底层界面面积的比例的量化测度需求日益提升。现有渗透率的测度主要依赖于成本高、效率低的手工分析，难以进行大规模、高效地测度。针对这一问题，本研究基于开源街景数据和机器学习算法，提出了一套人本视角下街道渗透率大规模、精细化测度和分析方法，并以上海中心城区为例，快速高效地实现了该范围内街道渗透率的计算和可视化。人工标注的结果与计算机的智能化识别在校核中显示了较高的拟合度，证明了该方法的有效性。实证分析发现，上海中心城区内街道渗透率存在明显的空间异质性，呈现“内高外低”的空间格局。本研究对经典城市设计要素与新数据新技术的深度整合方法可为人本导向的城市设计实践提供有力支持，同时兼具大规模与高精度的宏观图解也有助于提高设计师对街道空间的深入认知。
Abstract：Accompanying with the delicacy transformation of urban planning and design, humanscale street qualities have been regarded as key issues in recent years. Street transparency refers to the percentage of the area of the street door and window openings to the area of the street interface.Nevertheless, existed studies mainly rely on manual-based analytical approaches with high-cost and lowefficiency, which is hard to be measured on the city scale. As a response to this issue, this paper proposes a set of large-scale and refined measurement and analysis methods for street transparency based on the integration of street view images and deep learning. Taking Shanghai as an example, this paper completes the calculation and visualization of street interface transparency effectively. The verification via manual labeling obtains high coefficients with automatically computed results in statistical analysis, which proves the validity of this study. The empirical analysis finds that there is obvious spatial heterogeneity in street transparency in the central Shanghai, showing a spatial pattern of “high inside and low outside”. This paper is a result of integrating classical urban design concerns with new data and new techniques, which helps support human-oriented urban design practice; and it reveals a big picture co-presenting large-scale and high-precision results, which helps designers to seek in-depth understandings in this direction.
[1]唐婧娴,龙瀛.特大城市中心区街道空间品质的测度——以北京二三环和上海内环为例[J].规划师, 2017, 33(2):68-73.
[2]龙瀛,周垠.街道活力的量化评价及影响因素分析——以成都为例[J].新建筑, 2016(1):52-57.
[3]樊钧,唐皓明,叶宇.人本尺度下的社区生活便利度测度——基于多源城市数据的精细化评估[J].新建筑, 2020(5):10-15.
[4] EWING R, HANDY S. Measuring the unmeasurable:urban design qualities related to walkability[J]. Journal of urban design, 2009, 14(1):65-84.
[5] MONTGOMERY J. Making a city:urbanity, vitality and urban design[J].Journal of urban design, 1998, 3(1):93-116.
[6] EWING R H, CLEMENTE O, NECKERMAN K M, et al. Measuring urban design:metrics for livable places[M]. Washington, DC:Island Press,2013.
[7] GEHL J, KAEFER L J, REIGSTAD S. Close encounters with buildings[J].Urban design international, 2006, 11(1):29-47.
[8] EWING R, HANDY S, BROWNSON R C, et al. Identifying and measuring urban design qualities related to walkability[J]. Journal of physical activity and health, 2006, 3(supplement 1):S223-S240.
[9] HAMIDI S, MOAZZENI S. Examining the relationship between urban design qualities and walking behavior:empirical evidence from Dallas, TX[J].Sustainability, 2019, 11(10):2720.
[10] LóPEZ T G. Inf luence of the public-private border configuration on pedestrian behaviour:the case of the city of Madrid[D]. Spain:La Escuela Te c'nica Superior de Arquitectura de Madrid, 2003.
[11] City of San Jose. Downtown design guidelines[R]. San Jose:Planning Division of San Jose, 2004.
[12]陈泳,赵杏花.基于步行者视角的街道底层界面研究——以上海市淮海路为例[J].城市规划, 2014, 38(6):24-31.
[13] JACOBS A B. Great streets[J]. ACCESS magazine, 1993, 1(3):23-27.
[14] JACOBS J. The death and life of great American cities[M]. Vintage, 2016.
[15]徐磊青,康琦.商业街的空间与界面特征对步行者停留活动的影响——以上海市南京西路为例[J].城市规划学刊, 2014(3):104-111.
[16]吴莞姝,钮心毅.建成环境功能多样性对街道活力的影响研究——以上海市南京西路为例[J].南方建筑, 2019(2):81-86.
[17]叶宇,张昭希,张啸虎,等.人本尺度的街道空间品质测度——结合街景数据和新分析技术的大规模、高精度评价框架[J].国际城市规划,2019, 34(1):18-27. DOI:10.22217/upi.2018.490.
[18] YE Y, ZENG W, SHEN Q, et al. The visual quality of streets:a humancentred continuous measurement based on machine learning algorithms and street view images[J]. Environment and planning b:urban analytics and city science, 2019, 46(8):1439-1457.
[19] MIDDEL A, LUKASCZYK J, ZAKRZEWSKI S, et al. Urban form and composition of street canyons:a human-centric big data and deep learning approach[J]. Landscape and urban planning, 2019, 183:122-132.
[20] CAMPBELL A, BOTH A, SUN Q C. Detecting and mapping traffic signs from Google Street View images using deep learning and GIS[J]. Computers,environment and urban systems, 2019, 77:101350.
[21] LIU L, SILVA E A, WU C, et al. A machine learning-based method for the large-scale evaluation of the qualities of the urban environment[J].Computers, environment and urban systems, 2017, 65:113-125.
[22] QUERCIA D, PESCE J P, ALMEIDA V, et al. Psychological Maps 2.0:A web gamification enterprise starting in London[C]//Proceedings of ACM Conference on World Wide Web(WWW), 2013:1065-1076.
[23] QUERCIA D, SCHIFANELLA R, AIELLO L M. The shortest path to happiness:recommending beautiful, quiet, and happy routes in the city[C]//Proceedings of the 25th ACM conference on Hypertext and social media,2014:116-125.
[24] QUERCIA D, O’HARE N K, CRAMER H. Aesthetic capital:what makes London look beautiful, quiet, and happy?[C]//Proceedings of the 17th ACM conference on Computer supported cooperative work&social computing, 2014:945-955.
[25] NAIK N, PHILIPOOM J, RASKAR R, et al. Streetscore-predicting the perceived safety of one million streetscapes[C]//Proceedings of the IEEE conference on computer vision and pattern recognition workshops, 2014:779-785.
[26] WANG R, LIU Y, LU Y, et al. Perceptions of built environment and health outcomes for older Chinese in Beijing:a big data approach with street view images and deep learning technique[J]. Computers, environment and urban systems, 2019, 78:101386.
[27] CHEN L, LU Y, SHENG Q, et al. Estimating pedestrian volume using Street View images:a large-scale validation test[J]. Computers, environment and urban systems, 2020, 81:101481.
[28] LAUMER D, LANG N, VAN DOORN N, et al. Geocoding of trees from street addresses and street-level images[J]. ISPRS journal of photogrammetry and remote sensing, 2020, 162:125-136.
[29] YE Y, RICHARDS D, LU Y, et al. Measuring daily accessed street greenery:a human-scale approach for informing better urban planning practices[J].Landscape and urban planning, 2019, 191:103434.
[30] LI X. Examining the spatial distribution and temporal change of the green view index in New York City using Google Street View images and deep learning[J]. Environment and planning b:urban analytics and city science,2021, 48(7):2039-2054.
[31] GONG F Y, ZENG Z C, ZHANG F, et al. Mapping sky, tree, and building view factors of street canyons in a high-density urban environment[J].Building and environment, 2018, 134:155-167.
[32] TANG J, LONG Y. Measuring visual quality of street space and its temporal variation:methodology and its application in the Hutong area in Beijing[J].Landscape and urban planning, 2019, 191:103436.
[33] WANG R, LU Y, ZHANG J, et al. The relationship between visual enclosure for neighbourhood street walkability and elders’ mental health in China:using street view images[J]. Journal of transport&health, 2019, 13:90-102.
[34] ORDONEZ V, BERG T L. Learning high-level judgments of urban perception[C]//European conference on computer vision. Springer, Cham,2014:494-510.
[35] SRIVASTAVA S, VARGAS MUNOZ J E, LOBRY S, et al. Fine-grained landuse characterization using ground-based pictures:a deep learning solution based on globally available data[J]. International journal of geographical information science, 2020, 34(6):1117-1136.
[36]扬·盖尔.交往与空间[M].北京:中国建筑工业出版社, 2002:58.
[37] HURTIK P, MOLEK V, HULA J, et al. Poly-YOLO:higher speed, more precise detection and instance segmentation for YOLOv3[DB/OL]. arXiv:2005.13243v2, 2020[2021-01-29]. https://arxiv.org/abs/2005.13243.
[38] WU M, ZENG W, FU C W. Floorlevel-net:recognizing floor-level lines with height-attention-guided multi-task learning[J]. IEEE transactions on image processing, 2021, 30:6686-6699.
[39]胡军,张超,陈平雁.非参数双变量相关分析方法Spearman和Kendall的Monte Carlo模拟比较[J].中国卫生统计, 2008, 25(6):590-591.
(1)从室内看室外街道活动或者从室外看室内活动的视域范围和清晰程度。
(1)本文研究范围根据《上海市城市总体规划（2017-2035）》中的中心城区范围划定。
(1)该训练样本量参考了目前在自动驾驶和城市研究领域广泛应用的知名数据集Cityscape的数据量予以确定。Cityspace数据集由包括戴姆勒在内的三家德国单位联合提供，包含50多个城市的立体视觉数据，共包括5 000张精细标注图像，其中有2 975张训练图、500张验证图和1 525张测试图。本研究选择的3 500张满足深度学习的样本量，足以支撑识别的准确性。详见https://www.cityscapes-dataset.com/。
(2) Io U指数是用来检测算法运作是否良好的指标。Io U指数计算的是“预测的边框”和“真实的边框”的交集和并集的比值。其计算公式是：Io U指数=边框的重叠面积/（两个边框的总面积-重叠面积）。当Io U=1时，预测边界和实际边界完全重合。一般设定Io U阈值为0.5，值越大，边界框越精确，意味着识别效果越好。
(3)本次研究采用的百度云服务器是NVIDIA Tesla T4 8G。所租用的百度云服务器通过GPU加速后的识别街景图像的效率约2.9秒/张，所有街景图像识别总耗时390小时。
基本信息:
DOI：10.19830/j.upi.2021.241
中图分类号:TU984
引用信息:
[1]邵源,叶丹,叶宇.基于街景数据和深度学习的街道界面渗透率大规模测度研究——以上海为例[J].国际城市规划,2023,38(06):39-47.DOI:10.19830/j.upi.2021.241.
基金信息:
国家自然科学基金面上项目（52078343）; 国家重点研发计划（2023YFC3805503）
2022-02-15
2022-02-15
2022-02-15