Page 38 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

Crash_Report_Temporally_Related =
ℎ
ℎ
[ ℎ_time( ) ≥ Report_Open time( ) ∩ Crash_time( ) ≤ Report_Close time ( )] (1)
ℎ
ℎ

Where i and j represent the number of crashes (i =1, 2…,17360) and the number of reports (j=1, 2…,524525), respectively.
Spatial correlation: After applying the mentioned temporal correlation strategy on the dataset, the spatial correlation of
reports and accident data is examined. For this purpose, it is necessary to find an optimal spatial threshold (distance) to consider
the relevant reports to each accident. Indeed, for each type of crash, we use an optimal spatial threshold to filter some temporally
related reports that are not spatially related.
Land use of accident: to determine the land use of each accident; First, the dominant land use for each street segment was
determined. For this purpose, urban road data and parcel data were used. The main land use for each street segment is calculated
based on street segment’s land use density index. In this case we used a combination of two commonly used methods to calculate
land use density:
a) Counting: This index is defined based on the number of land uses placed on a road segment. For this purpose, the
percentage of the number of parcels of each land use to the total parcels in each street segment was used ([3], [21]).
b) Area ratio: This index is defined based on the ratio of the total area of different types of land uses. The method of
calculating the percentage of the area of each land use to the total area of land uses in each street segment was used
([3], [13]).
First, for each street segment the values of “frequency" and "area " are calculated separately for each land use (i =1,2,…,7).
Then, the obtained values for these two variables are normalized based on the "total frequency" and "total area" of the parcels
placed on each street segment. This repetitive process is performed for all of Boston's urban roads (Eq. 2 and 3).

( ) = ( ) × ( ) ، = 1،2، … ، 7 (2)

= ( ( )) (3)

While F(i) is the index value of each land use in each street segment and DLU is the dominant land use for each street
segment. Fig 2. shows a sample of the dominant land use in each street in a part of Boston. Fig 3. Visualized the road segment
and accidents based on the dominant land use assigned to each road segment.

Fig. 2. Example of Dominant Land Use

3