Using Google Street View to map the openness of street canyons of Boston

City streets are a focal point of human activity in urban centers. Citizens interact with the urban environment through its streetscape and it is imperative to, not only map city streetscapes, but quantify those interactions in terms of human well-being. Researchers now have access to fully digitized representation of streetscapes through Google Street View (GSV), which captures the profile view of streetscapes and, thus, shares equivalent viewing angles with those of the citizen. These two facets — a wealth of streetscape photographs at city-scale and a shared perspective with the end user — underscore the potential of these data in street-level urban landscape mapping. Here I present one of my recent work on using Google Street View data to map the openness of the street canyons in Boston, Massachusetts.

The sky view factor (SVF) is a geometric quantification of the openness or the degree of sky visibility within street canyons and has a wide array of applications as an indicator of urban form, structure, and proxy of localized environmental conditions. When the sky is totally obstructed in street canyons, the SVF is zero while the SVF is one when there is no obstruction.The SVF has been applied in studies of forestry, urban climate, urbanization, air pollution, and urban heat island effects. We have proposed to use the publicly accessible Google Street View panoramas to estimate the SVF at regional scale based on geometrical transform and image classification. We have collected over 11, 000 GSV panoramas for SVF estimation in Boston, Massachusetts (Fig. 1).

Fig. 1. The spatial distribution of the generate sample sites in Boston. Green sites represent those sites having Google Street View images captured in green seasons.

In order to calculate the SVF, we first created fisheye images by projecting those collected GSV panoramas from cylindrical projection to azimuthal projection. Fig. 2 shows the geometric model for the transformation of cylindrical projection to azimuthal projection.

Fig. 2. Transform of cylindrical GSV panorama to azimuthal fisheye image.

Fig. 3. GSV panorama and image classification result, (a) a cylindrical projection GSV panorama, (b) the generated azimuthal projection fisheye image, (c) the segmentation result, (d) the preliminary classification result using spectral rules, (e) the refined classification result after applying geometrical rules.

We further used an unsupervised object-based image classification method to segment sky pixels from the simulated fisheye images automatically (Fig. 3). The meanshift algorithm was used to segment the fisheye images (Comaniciu and Meer 2002). Based on the sky classification results, we applied the classical photographic method to calculate the SVF. The SVF is calculated as formula,

where, i is the ring index, n is the number of rings (here is set to 37), and αi is the angular width in ith ring. When the sky is totally obstructed, the SVF is zero while the SVF is one when there is no obstruction.

By applying the above process to all retrieved GSV panoramas along the city streets, I mapped the distribution of estimated SVF values in Boston at both the point level (Fig. 4) and the census tract level (Fig. 5). It can be seen clearly that the downtown area and the southwestern part of the study area have much lower SVF values compared with other parts of the city. The SVF value is determined by the obstructions of building block and tree canopies. In the downtown area, the obstructions of high-rise buildings lead to low SVF values or relatively closed street canyons.

Fig. 4. The spatial distribution of the Sky View Factor in Boston. Darker sites have higher level openness, and ligher sites have lower level openness.

Fig. 5. The spatial distribution of the SVF values in Boston, Massachusetts.

The openness of street canyons is an important geometrical parameter for the study of urban microclimate, air pollution migration, and human perception of the environment. Currently, a generalized and widely applied method to compute SVF in urban landscapes does not exist. Here, we demonstrated an application of GSV panoramas to, rapidly and fairly simply, estimate the street openness of Boston, Massachusetts. We have shown how a geometric transformation and simple image analysis can be used to automate a workflow to describe SVF across cities, wherein the only input is the publicly accessible GSV data. As such, the GSV-based method is suitable for large-scale SVF estimation can help researchers, urban planners and managers better understand the influence of urban form on the urban microclimate, urban air pollution migration, and human perception of urban environment.

Reference

Li, X., Ratti, C., & Seiferling, I. (2018). Quantifying the shade provision of street trees in urban landscape: A case study in Boston, USA, using Google Street View. Landscape and Urban Planning, 169, 81–91.

Li, X., Ratti, C., & Seiferling, I. (2017, July). Mapping urban landscapes along streets using google street view. In International Cartographic Conference (pp. 341–356). Springer, Cham.