3.1 Study Population

This project will cover all one of the five boroughs in New York City, Manhattan. The city is an ideal location for the research on greenery rate analysis and social injustice issues related to it, because it is one of the most populous and diverse cities in the world (>8.4 million people by 2019)(United States Census Bureau, 2018). And during pandemics, marginalized low-income groups tend to suffer more from both mental and physical health problems(NYC.gov, 2022).

​

3.2 Street View Images

We used Google Street View Static API to request the street view images as the test set. Google Street View Static API provides 360 panorama images based on location information across large regions of the United States. In the process of single image request and response, the view point is defined with URL parameters including location, fov, heading, pitch and radius and the street view picture is sent as a response through a standard Http request.(Google, n.d.)

​

The road network information was obtained through Open Street Maps dataset and includes all information under the “street” category. The test viewpoints were evenly selected from streetnet through 3D visual programming platform Rhino and its plugin, Grasshopper. The plugin Urbano was used to import GIS shapefile information to the Grasshopper environment. And the plugin Kangaroo was used to decrease overlapping points. We obtained street view points and its GIS location information was used to request for test street view images. In the practice, the images with heading of 0,90,180,270 and fov of 90 was selected.

​

3.3 social and neighborhood conditions current

To explore the current socio economic condition of neighborhoods, we downloaded a series of csv files with geoid from the US census database(U.S. Census Bureau, n.d.), including occupied housing units, median household income, monthly housing cost, total population, population of different age groups and enthnic groups, disability rate and education level(U.S. Census Bureau, n.d.).

​

3.4 machine learning model

We adopted a pre-trained model named efficient neural network. Its architecture was raised in 2016 by researchers in Mathematics, Informatics and Mechanics University of Warsaw and has the ability to perform pixel-wise semantic segmentation in real-time, up to 18× faster, requires 75× less FLOPs, has 79× less parameters, and provides similar or better accuracy to existing models without relying on a supercomputer (Adam Paszke,2016).

​

The ENet has the ability to distinguish the list of classes (Automatic Addison,2021):

Unlabeled, Road, Sidewalk, Building, Wall, Fence, Pole, TrafficLight, TrafficSign, Vegetation, Terrain, Sky, Person, Rider, Car, Truck, Bus, Train, Motorcycle, Bicycle.

​

Considering the amount of test data and hardware we have, we finally choose this algorithm.

​

3.5 Methodological framework

As shown in Figure 3, we developed our framework according to the literature review indicated in figure 3. The whole structure is divided into 2 parts. In the machine learning part, 2000 points were selected  from the street network in Manhattan(NYC Open Data, n.d.), and their location data was extracted in GIS as an URL parameters for google street view api to get street view images. Each viewpoint requested 4 pictures in four different directions. In the meantime, we used the Efficient Neural Network model to identify 2 kinds of street vegetation through pixel based semantic segmentation, bushes and trees. Later we imported all the images we had into the pre-trained model and exported the ratio of selected classes pixel by total pixel amount to calculate the greenery rate of each viewpoint.

​

And this leads to the analysis part, which is comparing greenery rate with multiple neighborhood features. The exported greenrt rate data was grouped and averaged based on census tract boundaries. The census average greenery rates were then merged with neighborhood features based on GEOID. We set LiDAR data for the comparison group to ensure the accuracy of the data. The result went well. Then through standardizing the data, filtering the no meaning data, fitting three different regression models and comparing the model performance, we managed to find out the correlations and significance level of each feature.

​

3.5.1 Image processing

To elaborate the analysis process in more detail, in order to obtain test grid matrix,  first we imported the shapefile of street network, extracted all street vertices as shown in figure 4, which are more than 10,000 points, which might take us a long time to calculate the greenery rate of 40,000 images, so later we used grasshopper to filter the data, deleted duplicated points and points that are too close to each other, which cut down the number of points to 2000, then we export all points with location data as stated in table 1.