Proposal of System for Calculating Sky View Factor Using Google Street View

Abstract

In recent years, big data entered use in various fields. Google Street View (hereinafter called “GSV”) can be regarded as open big data, and its images can be obtained using API. The streets can be viewed 360° horizontally and 290° vertically from each point on the web. In addition to those, zooming up is available, and the viewpoint can be moved approximately 10 m forward or backward to/from the current point. The original image to provide the view is the panoramic image associated with the latitude and longitude information on the street consecutively at intervals of 10 m, and they exist as massive data on the web. We determine the area of the sky using these images from GSV. In this research, we calculate the sky view factor (hereinafter called “SVF”) in an extended area by defining the area of the sky with the SVF and utilizing the computer.

This paper is recomposed according to the content of Nishio and Ito (2014).

1 Introduction

1.1 Background and Purpose

In accordance with the remarkable development of computers and the internet in recent years, big data entered use in various fields, including medical field, education, transportation, energy, housing, and urban development. Google Street View (hereinafter called “GSV”) ¹can be regarded as open big data, and its images can be obtained using API. The streets can be viewed 360° horizontally and 290° vertically from each point on the web. In addition to those, zooming up is available, and the viewpoint can be moved approximately 10 m forward or backward to/from the current point. The original image to provide the view is the panoramic image of the horizontal direction of 360° and vertical direction of 290°. Each image is associated with the latitude and longitude information on the street consecutively at intervals of 10 m, and they exist as massive data on the web. We determine the area of the sky using these images from GSV. In this research, we calculate the sky view factor (hereinafter called “SVF”) in an extended area by defining the area of the sky with the SVF and utilizing the computer.

The SVF was introduced as the concept of the easing of the restriction of the oblique line when Japan’s Building Standards Act was revised in 2003, which indicates the amount of the sky to the hemispherand in the equisolid angle projection (Fig. 22.1). Generally, the method with the software, such as JWCAD, is used to calculate the SVF. Furthermore, the method using the digital camera with the fisheye lens equipped to photograph the fisheye image is used in previous researches (Takei and Ohara 1977; Ito et al. 2005). These methods are effective for calculating the SVF accurately at a single point. On the other hand, Ministry of the Environment (2003) calculates the SVF by modeling the building arrangement within the 500 m mesh. These methods are effective when understanding the distribution of the SVF in an extended area.

Fig. 22.1

Definition of SVF

As explained above, there are a variety of methods to calculate the SVF; however, none of them satisfy both conditions, in detail and in an extended area, to calculate the SVF at sequential points on the streets. Therefore, we suggest the system in this paper, which calculates the SVF automatically by converting the panoramic images from GSV to the fisheye images. The purpose of this study is to suggest the new system which allows the calculation of the SVF both in detail and in an extended area.

Acquiring the SVF in an extended area and in detail makes it easier to compare and analyze it with the psychological indicator, such as a feeling of oppression on the streets, and the forming, characteristics, and form restriction of the streetscape in an extended area. Moreover, we believe that releasing this system makes it possible for everyone to calculate the SVF anywhere within the photograph points of GSV, and there is a possibility of the development from the perspective of the big data utilization.

1.2 Definition of SVF

1.2.1 Definition of SVF in This Paper

The definition of the SVF in this paper is the amount of the visible sky to the whole sky at arbitrary measurement point shown in the equisolid angle projection. There are two differences from the general definition of the SVF. The first one is what is regarded as the “visible sky.” Whereas it is the ratio of the sky in the condition with only the building to be constructed in Japan’s Building Standards Act, it is the ratio of the visible sky to the whole sky in the condition with all buildings, planting, telegraph poles, and signs in this research. The second one is the measurement point. Because the SVF is calculated from GSV, the measurement point is the center of the road at the height of 2.45 m or 2.05 m. As described above, the SVF which is dealt with in this research is different from the one in Building Standards Acts in Japan: however, we use it following other previous researches.

1.2.2 SVF Calculation Using the Fisheye Lens

We compared the SVF calculated from GSV with the one calculated using the fisheye lens in order to verify the consistency. The images were photographed using the converter (NikonFisheyeConverterFC-E8) attached to the compact digital camera (NikonCOOLPIX4500) and the tripod (SLIKPRO330DX II) in the equidistant projection at each measurement point. They were converted to the orthograph using SPCONV (n.d.) to create fisheye images (Fig. 22.2). The SVF calculated with the above steps is defined as true.

Fig. 22.2

Example of fisheye image

2 Method for Calculating SVF

2.1 Overview of Calculation Method (Fig. 22.3)

This system consists of two programs: (1) a program which acquires the latitude, longitude, and PanoID information from GSV and (2) a program which acquires and saves the corresponding panoramic image with PanoID and calculates the SVF with the image processing (Fig. 22.3). PanoID is the number which is given to each 360° panoramic image of GSV. We implemented the program (1)² to the web browser (http://www.comp.tmu.ac.jp/fiweb/itoken/nishi/panoindex.html), which can acquire and generate the list of the PanoID, latitude, and longitude information of the point where the panoramic image was photographed in an extended area. The program (2) then acquires the corresponding panoramic image (mercator projection) with PanoID, performs the edge detection with the image processing, and creates the binary image of the area of the sky and the one of the others. In the next place, the image was converted in the orthograph, and the ratio of the visible sky to the whole sky was calculated. Automatizing these steps makes it possible to calculate the SVF at intervals of 10 m in an extended area.

Fig. 22.3

Overview of SVF Calculation System

2.2 Creation of List of PanoID, Latitude, and Longitude

Firstly, the following specifications were implemented on the web browser (http://www.comp.tmu.ac.jp/fiweb/itoken/nishi/panoindex.html) using Google Maps API V3 within HTML by JavaScript. The latitude, the longitude, and PanoID are described with a focus on it at the bottom of the browser screen when arbitrary point is clicked (Program 1 in Fig. 22.3). Additionally, the specification is implemented which repeats the description of the GSV points automatically and continuously. The point whose information was already acquired was stored to the memory in order to avoid acquiring it from the same point. The latitude, longitude, and PanoID were described in each sentence at the bottom of the browser screen and copied to the text file for further analysis. It is noted that according to the limitations on the use of Google Maps API, the number of times the images can be read is restricted to up to 25,000 times per API per day. The images should be obtained within the limitation.

2.3 Method for Extracting Sky by Image Processing

The panoramic image (512 px × 512 px) was acquired, and the area necessary for the SVF calculation (416 px from the top left to the right and 110 px to the bottom) was cut automatically using Python in order for each PanoID. The panoramic image of the point (RGB color) was converted to the grayscale image to make the image processing easy. The edge detection was then performed. The edge was dilated, and a process called closing, which erodes the dilated edge, was conducted because it is significant for this edge detection to prevent the disconnection of the edge between the areas of the sky and the others. As seen in Fig. 22.4, the liner breaks were connected, the edge between the areas of the sky and the others was detected, and the binary image was created to detect the sky ((1) and (2) of Program 2 in Fig. 22.3).

Fig. 22.4

Closing (enlarged image)

2.4 Conversion of Projection Method and Calculation of SVF

The panoramic image was photographed in the mercator projection. The projection method of the fisheye image is the orthographic projection, where the image displayed on the hemispherand is projected longitudinally. We defined the length of the panoramic image as h, the width of it as w, the height of the cylinder as h, and the circumference of it as w.

The conversion equation from the panoramic image to the fisheye image is described as follows. We defined an arbitrary coordinate on the panoramic image as (X, Y), the corresponding point on the fisheye image as r·sinθ in the polar coordinate, and the radius of the hemispherand as R. The following relation is confirmed between the combination of X and Y and that of r and θ.

The SVF was calculated by counting the number of the pixels in the sky area (white) to the one in the whole area (circle) on the fisheye image converted using the above equation. For the above steps, the list output of the latitude, the longitude, PanoID, and the SVF for the acquired data was obtained ((3), (4), and (5) of Program 2 in Fig. 22.3).

2.5 Analysis of Consistency

The consistency of the results of the proposed system is verified visually and numerically. Firstly, it is visually verified whether the fisheye images created using the system proposed in this paper (example in Fig. 22.5) coincide with the ones photographed with the fisheye lens (example in Fig. 22.6). While there seems to be minor differences, the fisheye images are consistent overall. The minor differences include the smaller extension of the buildings toward the center compared to the fisheye image photographed with the fisheye camera. It is considered that there may be an issue with the height of the cylinder, which we defined, and we would like to work on it. Secondly, we verified with the values. When we compare the values between the SVF calculated from the fisheye images that are photographed with the fisheye lens and the one calculated with the system that is proposed in this study, the mean value of the errors between the two points with the SVF calculated is 1.2%, which is considered to be acceptable. It is assumed that the level of the errors varies from point to point. We aim to propose a more precise tool for the SVF calculation in future research by verifying the errors among points with the different factors, such as the shape, height, and width of the building, the road width, the building coverage ratio, etc.

Fig. 22.5

Created fisheye image

Fig. 22.6

Photographed fisheye image

3 Analysis of SVF Distribution in City

3.1 Creation of SVF Map

The SVF data which is associated with the latitude and longitude information is obtained in approximately 160 thousand points in the previous section, and it is imported to GIS and analyzed on the map. The data was obtained within 500 m of each station of 50 in total, 29 stations of JR Yamanote Line, 14 stations of JR Chuo Main Line, and 7 stations of JR Chuo Sobu Line, with their station building in the 23 wards. Among them, the points that appear to be unreliable are eliminated, and the SVF at approximately 130 thousand points is concerned in this section. The mean value of the SVF was 53.4%, the median was 51.1%, and the standard deviation was 13.6. As one of the examples, the distribution map of SVF around Shibuya Station is shown in Fig. 22.7.

Fig. 22.7

SVF Map centered on Shibuya Station

3.2 Use Districts and SVF

The figures, such as the floor area ratio and the height, are restricted based on the type of the use district. The relation between the use districts and the SVF is verified to clarify the difference of the SVF due to the level of the restriction. It is described in Fig. 22.8. Firstly, the high median and mean values of the SVF were confirmed in the quasi-residential districts among the residential districts. We consider it is because they are located along the relatively wide road. Secondly, the median and mean values of the SVF were higher in the commercial districts than the ones in the neighborhood commercial districts. We consider this is because of the tendency of the commercial districts being located along the arterial road, whose has the wide width. These results are obtained regardless of the fact that the commercial districts have lighter restrictions of both the use and form, and both the height and density of buildings are high. For the above reason, it is indicated that the SVF is especially related to the road width. Therefore, the relation between the road width and the SVF is analyzed in the next section.

Fig. 22.8

Box plot of SVF by use districts

3.3 Road Width and SVF

It is considered that the wider the road width is, the higher the SVF is. It is revealed that the mean value of the SVF of the points on the wide roads with the name³ is high compared to the one of all the points (53.4%) (Table 21.1). For instance, the one of 277 points on Ome-kaido Street is 60.2%, and the one of 281 points on Kanpachi-dori Street is 63.8% (Fig. 22.9). The quantitative analysis is set to be conducted in further research. For the reason, it is indicated that although high-rise buildings are located, it is possible to keep the SVF high by ensuring the road width sufficiently. According to Japan’s Building Standards Act, the floor area ratio and the restriction of the oblique line are defined as the form restriction by the use districts at the moment: however, we consider that the restriction and easing defined with the SVF are also effective.

Table 21.1 Mean SVF of wide streets

Fig. 22.9

Example of SVF on wide streets (Ogikubo Station Area)

3.4 Distance from the Station and SVF

We discuss the relation between the distance from the station and the SVF. No relation is revealed in terms of the points within 500 m of the station. Generally, there are many cases where high-rise buildings are often seen around the station, and more and more buildings with lower height and density are observed in accordance with the increase of a distance. Therefore, it is presumed that the bigger the distance from the station is, the higher the SVF is; however, such a tendency is not observed within 500 m. It appears that this is because the change point of the building use and arrangement is further than 500 m; therefore, we would like to expand the scope in further research.

4 Conclusion

The progress and use possibility regarding the construction of the system for calculating SVF are reported in this paper.

This system enables the SVF calculation at the points where GSV is disclosed. The generalization achieved by integrating subprograms and the accuracy improvement of the sky recognition regarding the image processing need to be addressed in the future.

The result of analyzing the SVF distribution in the 23 wards with this system suggests the tendency of a higher SVF at a wider road, which indicates the use possibility of the system.

The SVF at a number of points at intervals of 10 m is obtained by defining the panoramic image from GSV as big data. Although further analysis in detail is required, there is a possibility that the system will be substantially utilized for a new index of the urban planning and the town development according to the prospect of the restriction and easing with using the SVF.

We would like to improve the system to aim the elaboration and make it versatile and available for the public.