Hindawi

Mathematical Problems in Engineering

Volume 2022, Article ID 7891232, 14 pages
https://doi.org/10.1155/2022/7891232

Research Article
Analysis of Vehicle-Pedestrian Accident Risk Based on
Simulation Experiments

Rui Cheng ,1 Ye Pan ,1 and Lian Xie 1,2

1
Key Laboratory of ITS (Guilin University of Electronic Technology),
Education Department of Guangxi Zhuang Autonomous Region, Guilin 541004, China
2
Intelligent Transportation Systems Research Center, Wuhan University of Technology, Wuhan 430063, China

Correspondence should be addressed to Lian Xie; [email protected]

Received 16 May 2022; Revised 16 July 2022; Accepted 4 August 2022; Published 29 August 2022

Academic Editor: Meng Li

Copyright © 2022 Rui Cheng et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Vehicle-pedestrian accidents are one of the main types of road traffic accidents in China because of their mixed traffic features. By
analyzing the characteristics of vehicle-pedestrian accidents, the head injury criterion (HIC) was selected as a quantitative index of
pedestrian head injury risk, and vehicle-pedestrian collision simulation tests were carried out using PC-Crash. From the collected
test data, the multivariate relationship models between the HIC, vehicle speed, and collision angle were fitted for different vehicle
types. A risk assessment method for vehicle-pedestrian accidents based on the HIC was proposed by the Fisher optimal seg-
mentation algorithm. Finally, a new index for evaluating the accuracy of accident risk classification, the degree of error clas-
sification, was proposed to verify the validity of the accident risk assessment method. The results show that vehicle speed, collision
angle, and vehicle type play a key role in pedestrian injury. Flat-headed vehicles are more likely to cause head injuries to
pedestrians than high-headed and low-headed vehicles. Rear-end collisions cause more injuries to pedestrians than side collisions.
The research results can provide guidance and a basis for accident liability determination, speed limit management, vehicle safety
design, and human injury mechanism analysis.

1. Introduction Although most urban roads are divided into nonlane isolation
facilities, collision accidents between cars and pedestrians
With the rapid growth of the social economy and the con- cannot be eliminated.
tinuous innovation of science and technology, automobile The USA and the European Union began to pay at-
manufacturers are increasingly pursuing comfort, safety, and tention to pedestrian safety in the mid-1970s. Since then,
environmental protection, as well as the protection of pas- biological samples and mechanical dummies have been used
sengers in vehicles, while their awareness of the protection of to carry out real vehicle test studies instead of pedestrians,
pedestrians outside vehicles is relatively weak. Among all road including human injury evaluation, the impact of overall
traffic participants, pedestrians are one of the most vulnerable vehicle structure design on pedestrian injury, and human
groups. In recent years, the lack of safety awareness of drivers injury prevention measures and safety countermeasures
and pedestrians in observing traffic rules and poor traffic [2–5]. For example, Severy et al. used mechanical dummies
management has caused frequent pedestrian traffic accidents, to study vehicle-pedestrian collision accidents successively
and the resulting losses are difficult to estimate. According to in 1963 and 1966, which was the first experiment in this field
statistics in China’s annual statistical report of road traffic [6]. On this basis, other relevant scholars used dummies or
accidents (2021), pedestrian traffic accidents accounted for corpses to conduct a series of experiments under different
approximately 21% of all accidents in the whole year but lead collision conditions. The data obtained from a test are fitted
to approximately 27% mortality [1]. The reason is that road using mathematical regression with different methods to
traffic in China is mainly typical plane cross-mixed traffic. obtain a corresponding empirical formula [7].
2 Mathematical Problems in Engineering

Due to the high cost of real vehicle tests and considering model using ANNs and multiple logistic regression. By
the feasibility of computer technology simulation, some comparing the severity prediction results of pedestrian ac-
scholars have developed mature accident reconstruction cidents using different methods, they put forward the best
software (SMAC, AUTOSMAC, PC-Crash, etc.) by building prevention and control measures to improve pedestrian
digital dummy models, providing another way to analyze safety on rural roads [17].
and study vehicle-pedestrian collision accidents [8]. A In terms of vehicle and pedestrian characteristics, Park
dummy model should not only adapt to human body dy- et al. used a multilevel model to investigate the differences
namics analysis but also be able to carry out human body between the low-level individual characteristics and high-
damage analysis [9, 10]. Therefore, a digital dummy should level community environmental characteristics of pedestrian
have high computational efficiency, be able to adapt to collisions in Seoul, South Korea. The results showed that
different environments, conform to various biomechanical older pedestrians suffered more serious pedestrian injuries;
characteristics of the human body, and output corre- trucks and vans were more likely to cause serious pedestrian
sponding parameter information for the calculation of a injuries [18]. While researching the influence of alcohol,
damage index. Lasota’s survey found that younger victims were more likely
For example, Xu et al. proposed a personalized cus- to die at the scene of an accident, especially in rural areas
tomization method to divide the pedestrian structure into [19].
independent modules according to obvious bone markers To determine the high-risk and low-risk areas of pe-
and establish a multibody model and a finite element (FE) destrian death accidents in Iran, Hasani et al. input collected
model for each independent module to form a hybrid pe- pedestrian accident data into ArcGIS software to identify
destrian model [11]. To characterize the complex vehicle- high-risk and low-risk areas by calculating the spatial au-
pedestrian interaction process, Grindle et al. developed and tocorrelation of the data [20]. Jamali-dolatabad et al. col-
validated a detailed pedestrian FE model corresponding to lected pedestrian traffic accident data in Tabriz from 2014 to
50% male to predict injuries caused by pedestrian accidents. 2015, with fatal accidents as the case group and nonfatal
Compared with a simplified pedestrian model, the con- accidents as the control group. Their results showed that the
structed model shows higher biofidelity [12]. In addition, to significantly related factors influencing pedestrian death
explore the damage prediction ability of the Total Human were pedestrian age, type of license plate, accident season,
Body Model for Safety (THUMS) finite element human type of driving license, gender of pedestrian, and pedestrian
model (FE-HBM) in real-world vehicle-pedestrian colli- fault [21]. Feng et al. used a multivariate logistic regression
sions, Panday et al. used sequences of multibody tools and model to analyze 111 collision accidents and found that
finite element tools to reconstruct 10 vehicle-pedestrian collision with the front windshield frame usually leads to
crashes with lower limb injuries. The conclusion shows that more serious damage. When the collision speed exceeds
the THUMS FE-HBM can better predict pedestrian injuries 40 km/h, the risk of serious head injury for pedestrians
in real traffic accidents [13]. Lalwala et al. also reconstructed increases sharply [22]. While discussing the characteristics
a THUMS pedestrian model using pedestrian accident cases. of fatal pedestrian accidents involving low-speed vehicles,
It can be observed from the reconstruction study that the Matsui, a Japanese scholar, thought the relative percentage
kinematic response and damage response of their THUMS of pedestrian deaths is significantly higher when vehicles are
lower limb model are in good agreement with the actual traveling at low speeds, except for accidents involving box
collision data [14]. trucks or SUVs [23].
With the continuous improvement of accident databases At present, researchers mainly carry out vehicle-pe-
by traffic research institutions in various countries, in recent destrian accident research through simulation and accident
years, some scholars have relied on cases in a database to case statistical analysis. In terms of simulation, most existing
explore the significant cause factors affecting the frequency studies are optimized and improved for a dummy model,
and severity of vehicle-pedestrian accidents. For example, in and few involve the risk assessment of pedestrian accidents
terms of road control, based on three years of traffic accident by developed dummy models. However, an accident risk
data in Hong Kong, Zhu selected the best performance assessment based on the statistical analysis of accident cases
artificial neural network (ANN) model by comparing var- mainly considers macro factors, such as vehicle speed, ve-
ious data mining algorithms and determined the most hicle type, accident environment, and road design, but lacks
significant factors causing fatal and serious accidents. The in-depth discussion of the impact of micro-factors (such as
results show that in rainy weather, fatal and serious vehicle- pedestrian speed, collision angle, type of head, and pedes-
pedestrian collisions are more likely to occur in the case of trian and vehicle lateral contact position) on human injuries.
intersection signal failures [15]. Sheykhfard et al. used a Therefore, this paper adopts computer simulation tech-
structural equation model (SEM) to investigate the data of nology to carry out a risk analysis of vehicle-pedestrian
1358 pedestrian accidents in Gilan Province, Iran, from 2012 accidents with a more mature multirigid body dummy
to 2018. Factor analysis results showed that the quality model. Common computer simulation software for accident
defects of automobiles and poor traffic design at intersec- reappearance includes SMAC and AUTOSMAC, developed
tions, major urban roads, and outer ring roads are important by the National Road Traffic Safety Administration of the
reasons for the increase in fatal accidents [16]. With the USA, and PC-Crash, developed by the DSD of Austria. Since
continuous improvement of research methods, Kamboozia SMAC and AUTOSMAC cannot realize pedestrian collision
et al. constructed a pedestrian accident severity prediction simulation, PC-Crash has been constantly updated and
Mathematical Problems in Engineering 3

upgraded, adding pedestrian multirigid body and FE cal-

culation models on the basis of a single rigid body model,
which has been widely recognized in the field of road traffic
accident reconstruction and has become the most widely
used accident simulation software. In view of this, this paper
chooses PC-Crash to carry out simulation tests. Through the
investigation and analysis of accident cases, the typical Figure 1: Diagram of the collision process.
collision patterns of vehicle-pedestrian traffic accidents are
determined. Based on a rigid body dynamics method, PC-
Crash is utilized to build three vehicle rigid body models any object and only receives friction resistance from the air.
(short head, high head, and flat head) and a pedestrian ③ Landing and rolling stage: After falling to the ground,
multirigid body model, HTC is used to measure the risk of pedestrians are affected by the friction resistance of the road
pedestrian head injuries, a vehicle-pedestrian collision and immediately begin to slow down. According to the
simulation test is carried out under the influence of different different falling speeds and directions, the pedestrian ex-
vehicle types, vehicle speeds, pedestrian speeds, collision periences a composite motion state of sliding, rolling, or
angles, and relative collision positions, and the influence of rotating on the road surface.
multiple factors on pedestrian injury is studied. According
to the collected test data, the relationship model between the
HIC and vehicle speed under different collision angles is 2.2. Pedestrian Injury Mechanism. In vehicle-pedestrian
fitted using multiple regression analysis for different vehicle collision accidents, the injuries suffered by pedestrians can
models. A reasonable evaluation series and corresponding be roughly divided into two categories, namely, primary
HIC threshold of pedestrian injury risk in vehicle-pedestrian injuries and secondary injuries. A primary injury refers to an
accidents are determined using the Fisher optimal seg- injury caused by the first collision or crushing of the human
mentation algorithm, and the vehicle-pedestrian accident body between a vehicle and a pedestrian. A secondary injury
risk evaluation method based on HIC is given. Compared refers to an injury caused by the collision and scraping
with risk assessment based on accident case analysis, the between the human body and the ground or other objects
research results obtained by simulation in this paper are not after being hit. Generally, pedestrian injury mechanisms can
affected by factors, such as region, time, and traffic char- be divided into the following two categories according to
acteristics, and have universal applicability. different vehicle types:
(1) When a pedestrian collides with a high-head vehicle
2. Characteristics Analysis of Vehicle- (such as an off-road vehicle) or a low-head vehicle
Pedestrian Collision Accidents (such as a car), because the front collision contact
point of the vehicle is not higher than the center of
2.1. Collision Form. Compared with people in a vehicle, gravity of the human body, the first damage to the
pedestrians are completely exposed to the external envi- human body is the car’s front bumper, engine cover,
ronment, have strong subjective randomness when walking, and front windshield contact collision with the human
and choose the road, direction, and walking speed that they body. The second collision is the human body in the air
think will reach their destination. The characteristics of after the overturn of the vehicle, subsequent contact
pedestrian traffic are related to the height, age, psychology, collision or landing, and road impact causing damage.
physique, and gender of pedestrians, which leads to the
(2) When a pedestrian collides with a flat-headed vehicle
diversity and complexity of vehicle-pedestrian collision
(such as a van or truck), the height of the front part of
accident patterns and makes them face different injury risks.
the vehicle is higher than the height of the pedes-
There are three common collision patterns between cars
trian’s center of gravity, and the pedestrian’s legs,
and pedestrians: front, side, and rear collisions with pe-
torso, or even their whole body contact with the front
destrians. Because frontal collisions between vehicles and
part of the vehicle, resulting in a large contact surface
pedestrians occur most frequently and are most likely to lead
and easily causing nonobvious trauma to the human
to pedestrian deaths, this paper focuses on frontal collisions
body but especially serious internal injuries. After
between vehicles and pedestrians. This collision process can
the collision, the human body experiences a flat
be divided into three stages, as shown in Figure 1.
throwing movement, which will cause obvious
① Collision phase: The pedestrian’s lower limbs first
falling injury and then scratching or rolling.
collide with the front part of the vehicle. As the vehicle
continues to move forward, the pedestrian’s chest and ab- According to the statistical analysis results of pedestrian
domen contact and collide with the engine hood of the traffic accidents in China [24], Figures 2 and 3 show the
vehicle. If the vehicle speed is high, the pedestrian’s head will injury frequency of pedestrian body parts and the distri-
collide with the front windshield or A-pillar of the vehicle. bution frequency of fatal body parts, respectively. Figure 2
Usually, at this stage, pedestrians are subjected to a large shows that the most vulnerable parts are people’s legs and
impact force, resulting in very serious pedestrian damage. ② head, among which leg injuries account for 40%, and 32% of
Flight phase: After the collision, the human body continues pedestrians suffer head injuries. Figure 3 shows that head
to fly in the air. At this stage, the pedestrian does not contact and chest injuries are the main causes of pedestrian deaths,
4 Mathematical Problems in Engineering

head where a represents the resultant acceleration at the center of

32% gravity of the head in a collision; t1 and t2 are two different times
during the collision, which are chosen to maximize the HIC; and
1/t2 − t1 t2 at dt represents the average resultant acceleration
t
1
between t1 and t2. The time interval (i.e., from t1 to t2) of the HIC
substantially affects the calculation of its value. Thus, this paper
hand
selects 15 ms, as adopted by the U-NCAP collision test.
leg 12%
40%
3.2. Fisher Optimal Segmentation Method. When the
hips HIC > 1,000, head injuries tend to be more severe, but there is
other chest 10% no clear standard for using the HIC to measure the severity of
2% 4% head injuries. Therefore, this paper uses the Fisher optimal
segmentation algorithm to classify the HIC data obtained from
Figure 2: Distribution of pedestrian injuries.
a simulation test to obtain a pedestrian accident risk classifi-
cation scheme, including the optimal classification number
and corresponding index thresholds of each level. The principle
head
64%
of Fisher optimal segmentation is to ensure the ordered sample
data of each group after segmentation and minimize the sum of
squares of deviations within the group. In this case, the cor-
responding grouping is the optimal segmentation. The specific
hip implementation method is described below [26].
5%

3.2.1. Define and Calculate the Class Diameter. The Fisher

chest optimal segmentation algorithm usually uses the diameter to
12% define the difference degree in a class. When the difference
degree in the class is smaller, the class diameter is smaller,
other abdomen
8%
indicating that the sample attributes in the class tend to be
11%
consistent. If the data samples are divided into k categories
Figure 3: Distribution of fatal pedestrian parts due to injury. and have Ck−1m−1 classification methods, assuming that Hij is
one of the classifications, then the sample data contained in
the classification are denoted as xi , xi+1 , . . . , xj  (1 ≤ i < j).
accounting for 64% and 12%, respectively. Therefore, the
The sum of the squares of deviations of the samples in Hij is
analysis and protection of head injuries play an important
defined as the class diameter D (i, j), namely,
role in pedestrian safety research.
j
D(i, j)   xi − xij  xi − xij ,
T
(2)
3. Evaluation Standard and Method of i1
Head Injury
where xi is the standardized sample value and xij is the
3.1. Head Injury Criterion. The head injury evaluation average value from sample i to sample j.
standard is defined by physical parameters related to injury
intensity or a function composed of several physical pa- 3.2.2. Calculate the Classification Error Function.
rameters, which is related to the degree of risk of a certain Generally, an error function is used to define the merits and
part of the human body being damaged and is used to demerits of data sample classification. If m data samples are
measure whether the load exceeds the degree of causing a divided into k categories, the corresponding error function of
certain injury. At present, the head injury criterion (HIC) this category is
proposed by Versace in 1971 is widely adopted interna-
k
tionally, serves as the basis of the Federal Motor Vehicle
Safety Standards (FMVSS), and is used to evaluate vehicle e[p(m, k)]   Diη , iη+1 − 1. (3)
η1
safety. In domestic and foreign laws and standards, the HIC
has become the most extensively employed criterion for That is, the error function of any classification method is
evaluating head injuries. The safety limit value for the HIC is expressed as the sum of all diameters. The smaller e[p(m, k)]
generally 1,000. According to related statistics, when the is, the smaller the sum of diameters of all classifications is,
HIC > 1,000, the probability of fatal head fracture exceeds and the better the classification effect is.
33%. Its calculation formula is as follows [25]:
2.5
HIC  
 t2 − t1   a(t)dt 
1 t2
3.2.3. Determine the Optimal Solution. According to for-
(1)
t2 − t1 t1 max
mula (3), when k  2, the error function corresponding to the
optimal 2 classifications is
Mathematical Problems in Engineering 5

e[p(m, 2)] � min {D(1, i − 1) + D(i, m)}. Table 1: AIS classification standard.
2≤i≤m
(4)
AIS score Description Mortality rate (%)
When k > 2, the error function corresponding to the 1 Mild 0.6
optimal k classification is 2 Moderate 3.2
3 Heavy 9.3
e[p(m, k)] � min 􏼈e[p(i − 1, k − 1)] + D(i, m)􏼉, (5) 4 Severe 28.3
2≤i≤m
5 Critical 78.4
Here, it is necessary to determine the appropriate classifi- 6 Extreme (currently incurable) 100
cation point ik guarantee formula (5) to calculate the
minimum value, that is, e[p(m, k)] � e[p(ik − 1, k − 1)] +
D(ik , m) is the minimum, so the k class
Hk � 􏼈ik , ik+1 , . . . , m􏼉 can be obtained. Then, the classifica-
tion point ik-1 is determined to make it satisfy e[p(ik − 1, k −
1)] � e[p(ik−1 − 1, k − 2)] + D(ik−1 , ik − 1) so that the k−1
class Hk−1 � 􏼈ik−1 , ik−1 + 1, . . . , ik − 1􏼉 can be obtained, all
the classifications H1 , H2 , . . . , Hk can be obtained by Figure 4: Road scenario.
analogy, and finally, the optimal solution can be obtained.
an ellipsoid, which simplifies the joint of each part to a hinge
3.2.4. Determine the Optimal Classification Number. connection. In this paper, adult males were selected as the
Generally, the optimal classification is determined by research object, and multiple rigid body model parameters
drawing the curve of the minimum error function changing were assigned according to physical characteristics. The
with the classification number k. When the curve curvature specific setting parameters are shown in Table 2.
changes significantly, the corresponding k value is the ap- According to the analysis in Section 2.2, a Volkswagen
propriate classification number. In addition, to determine Bora 2.0, Audi Q5 2.0TDI, and Volkswagen T4 2.5 TDI (as
the k value more accurately, the minimum error function shown in Figure 6) were selected from the PC-Crash da-
ratio β (k) between adjacent classification numbers can be tabase as representative models of low-head, high-head, and
further calculated. The larger β (k) is, the better the clas- flat-head vehicles, respectively, to explore the impact of
sification effect is, as shown in different head types on pedestrian injury risk.
e[p(m, k)]
β(k) � . (6)
e[p(m, k + 1)] 4.2. Test Scheme. Considering that pedestrian speed may
affect accident risk, this research carries out vehicle-pe-
destrian collision simulation tests under a pedestrian
3.3. Abbreviated Injury Scale. In addition, to standardize the moving state. According to the statistical analysis of pe-
evaluation of injury levels after accidents, countries around destrian motion state and collision angle in vehicle-pedes-
the world have put forward the abbreviated injury scale trian accidents in the literature [27], 55% of pedestrians are
(AIS) to evaluate the risk of injury after years of accident in a walking state, 38% are in a running state, and 4% are in a
research. The AIS is a scoring method for classifying trauma stationary state. The proportion of pedestrians in the lateral
based on anatomical indexes. It defines nine anatomical position (i.e., the walking direction of pedestrians is 90° from
ranges to determine the location of trauma, specifically the driving direction of vehicles) is 68%, followed by the rear
including head (skull and brain), face (eyes and ears), neck, position (i.e., the walking direction of pedestrians is 0° from
chest, abdominal and pelvic organs, spine (cervical spine, the driving direction of vehicles) and the facing position (i.e.,
thoracic spine, and lumbar spine), upper limbs, lower limbs, the walking direction of pedestrians is 180° from the driving
pelvis and buttocks, body surface (skin) and thermal injury, direction of vehicles), accounting for 21% and 7%, respec-
and other injuries. According to the pedestrian injury report, tively. Therefore, this research chooses two states of walking
the trauma level of the pedestrian after the accident was and jogging for pedestrians (corresponding to a walking
assessed with six grades, as shown in Table 1. speed of adult males being approximately 5 km/h and a
jogging speed being approximately 10 km/h), and the two
4. Data Collection orientations of collision angle between pedestrians and cars
are 90° and 0° for experiments. In addition, to explore the
4.1. Test Scenario. First, a two-way four-lane road model impact of pedestrian collision position relative to vehicle on
with a lane width of 3.5 m, total length of 100 m, and accident risk, this study selects the front 1/4 and 1/2 posi-
pavement adhesion coefficient of 0.7 is constructed using tions as research variables. Figure 7 shows the collision
PC-Crash, as shown in Figure 4. Then, the software’s human position of pedestrians relative to cars at different collision
body model is used. The model is a multibody dynamic angles.
system consisting of 16 independent rigid bodies and 15 Considering the demands of road traffic car speed in
hinge joints, as shown in Figure 5. Each part of the human China, at the same time to make the research more rea-
body (head, trunk, limbs, buttocks, etc.) is regarded as an sonable, the car speed is set to VS ∈ {20, 30, 40, 50, 60, 70, 80,
independent rigid body, and its surface shape is defined by 90, 100, 110} km/h. Combined with the analysis of the above
6 Mathematical Problems in Engineering

Head

Neck

Upper arm
Trunk
Forearm
Hip

Thigh

Knee joint

Calf

Foot
Figure 5: Pedestrian multibody model.

Table 2: Body parameters of 90th percentile adult males aged 18–60 years in China.
Height (mm) Weight (kg) Upper arm length (mm) Forearm length (mm) Thigh length (mm) Calf length (mm)
1754 71 333 253 496 396

(a) (b) (c)

Figure 6: Test vehicle models. (a) VW Bora 2.0. (b) Audi Q5 2.0TDI. (c) VW T4 2.5 TDI.

Vp Vp

90˚ 90˚
Vs Vs

1/2W 1/4W

Vs Vs

Vp
Vp

1/2W 1/4W

Figure 7: Diagram of the test collision location.

Mathematical Problems in Engineering 7

Table 3: Resultant head acceleration value of low-head vehicle and pedestrian collision (m/s2).
1/4W 1/2W
Speed (km/h) 5 km/h 10 km/h 5 km/h 10 km/h
0° 90° 0° 90° 0° 90° 0° 90°
20 7325 4910 10743 5710 8893 5523 7418 5661
30 18432 8492 19722 5691 18817 7517 19366 5863
40 31446 10111 28788 12198 31143 7824 31050 9156
50 38902 16472 39020 12151 38179 17399 36127 11671
60 37538 19569 37202 20320 34151 26587 36897 17693
70 39112 29179 40570 25684 38536 32482 37900 32015
80 83079 57019 96861 46872 73810 36515 96522 30526
90 61532 70732 54079 52682 91077 58322 67637 48716
100 66272 70748 75818 58711 72905 54236 77429 35124
110 98648 73310 112221 75050 94712 55597 117232 60774

Table 4: Resultant head acceleration value of high-head vehicle and pedestrian collision (m/s2).
1/4W 1/2W
Speed (km/h) 5 km/h 10 km/h 5 km/h 10 km/h
0° 90° 0° 90° 0° 90° 0° 90°
20 6850 6018 4975 3768 9324 4959 9367 4932
30 19676 6205 21291 5873 17794 8436 18627 5697
40 27784 10803 28913 11685 26035 11672 26990 10196
50 41633 17836 39939 14931 34615 16708 35541 13995
60 30945 22409 32566 16596 33216 24322 32900 17927
70 41164 27664 37652 24207 57948 33650 45046 23883
80 70486 32955 60383 31991 118726 37902 89432 37764
90 82572 54328 69229 38435 69090 53954 59607 45422
100 78490 57858 91205 51966 69118 54485 136390 64674
110 96015 64113 160570 84012 136823 54876 112009 54378

Table 5: Resultant head acceleration value of flat-head vehicle and pedestrian collision (m/s2).
1/4W 1/2W
Speed (km/h) 5 km/h 10 km/h 5 km/h 10 km/h
0° 90° 0° 90° 0° 90° 0° 90°
20 72581 68028 57022 56185 102776 58738 96927 63485
30 55403 50536 65902 56045 93834 51304 77411 49532
40 63699 39733 84445 46869 76342 48448 86677 49439
50 78634 51023 103047 44617 115872 46188 102369 42441
60 115818 78840 130771 64405 152051 69207 147071 53681
70 130897 121951 168717 109352 157615 82202 171708 73181
80 154287 158591 144967 143215 207197 122008 206549 104993
90 152299 173736 147755 147764 183070 122477 177983 141675
100 282092 178430 244815 166398 264413 156806 260346 153319
110 375455 229065 418680 187684 430182 178198 448000 166874

variables, we set the model M ∈ {M1, M2, M3}, wherein M1, 5. Results and Analysis
M2, and M3 represent the low-head, high-head, and flat-head
vehicles, respectively; collision angle θ∈ {0°, 90°}; pedestrian 5.1. Model Fitting. Through each simulation test, the re-
walking speed VP ∈ {5 km/h, 10 km/h}; collision position R ∈ sultant acceleration value of the pedestrian head under
{1/4 W, 1/2 W}, where W is the width of the front part of the different collision forms is obtained, as shown in Tables 3–5.
car in meters. The simulation test involves 5 variables, namely, The HIC value can be calculated using formula (1), as shown
{M, θ, VS, VP, R}. According to the values of each variable, a in Tables 6–8. Among them, the test data with a HIC val-
total of 3 × 2 × 10 × 2 × 2 � 240 collision modes were simu- ue ≥ 1000 account for 75.8% of the total.
lated. Before the collision, the vehicle moves in a straight line, First, the Pearson correlation analysis between each
and after the collision, the vehicle is fully braked until it stops. variable and the HIC is carried out, and the results are listed
The collision process is shown in Figure 1. in Table 9. ∗ ∗ in the table represents a significant
8 Mathematical Problems in Engineering

Table 6: HIC value of collision head between low-head vehicle and pedestrian.
1/4W 1/2W
Speed (km/h) 5 km/h 10 km/h 5 km/h 10 km/h
0° 90° 0° 90° 0° 90° 0° 90°
20 46 17 121 25 75 23 48 24
30 466 67 551 25 490 49 527 27
40 1770 104 1419 166 1728 55 1715 81
50 3013 352 3036 164 2875 403 2504 149
60 2756 541 2695 594 2176 1164 2640 420
70 3054 1468 3347 1067 2943 1920 2823 1851
80 20084 7837 29477 4802 14942 2572 29220 1644
90 9481 13432 6866 6431 25272 8293 12011 5288
100 11414 13440 15979 8432 14488 6916 16842 2334
110 30856 14690 42589 15577 27870 7358 47504 9192

Table 7: HIC value of collision head between high-head vehicle and pedestrian.
1/4W 1/2W
Speed (km/h) 5 km/h 10 km/h 5 km/h 10 km/h
0° 90° 0° 90° 0° 90° 0° 90°
20 39 28 18 9 85 17 86 17
30 548 31 668 27 426 66 478 25
40 1299 122 1435 149 1104 149 1208 106
50 3570 429 3218 275 2250 364 2404 234
60 1701 759 1932 358 2030 931 1982 434
70 3471 1285 2777 920 8160 2097 4348 890
80 13316 1990 9045 1848 49033 2823 24146 2798
90 19778 6945 12730 2924 12666 6826 8757 4439
100 17424 8129 25361 6215 12679 6995 69354 10738
110 28838 10507 104299 20653 69906 7122 42389 6961

Table 8: HIC value of collision head between flat-head vehicle and pedestrian.
1/4W 1/2W
Speed (km/h) 5 km/h 10 km/h 5 km/h 10 km/h
0° 90° 0° 90° 0° 90° 0° 90°
20 14327 12185 7838 7554 34185 8441 29527 10251
30 7294 5796 11255 7507 27228 6019 16832 5512
40 10338 3177 20919 4801 16256 5216 22329 5486
50 17504 5936 34412 4245 46139 4629 33848 3746
60 46085 17619 62430 10627 91011 12720 83740 6740
70 62581 52429 118035 39919 99565 19558 123337 14626
80 94393 101114 80777 78359 197276 52491 195738 36059
90 91382 127011 84717 84730 144763 52997 134915 76269
100 426669 135763 299372 114021 362929 98293 349136 92919
110 871989 253521 1145038 154059 1225307 135322 1356163 114839

Table 9: Correlation analysis.

Vehicle speed Collision position Pedestrian speed Collision angle Vehicle type
HIC 0.333 ∗∗ 0.017 0.005 −0.181 ∗∗ 0.311 ∗∗

correlation at a level of 0.01 (two-tailed test). The vehicle shows that the impact force of any part of the front of the
speed, collision angle, and vehicle type are significantly vehicle on the pedestrian is equivalent, and the pedestrian
correlated with the HIC, while the collision position and speed relative to the vehicle speed can be ignored at the
pedestrian speed show a weak correlation. The analysis moment of collision. The above reasons lead to a change in
Mathematical Problems in Engineering 9

1400000 1400000

1200000 1200000

1000000 1000000

800000 800000

HIC
HIC

600000 600000

400000 400000

200000 200000

0 0
20 40 60 80 100 120 0 45 90
Vehicle speed (km/h) Collision angle (˚)

Figure 8: Relationship between vehicle speed and the HIC. Figure 10: Relationship between impact angle and the HIC.

Table 10 shows how scholars use various methods to

identify the significant risk factors based on data sources
1400000 from different areas and times. Besides collision angle, the
significant risk factors identified in this paper are basically
1200000
consistent with those identified by other researchers, such as
1000000 the vehicle speed and vehicle type. In terms of vehicle type,
factors screened by other scholars, such as larger vehicles,
800000 light machinery trucks, and light passenger vehicles, all
HIC

belong to the category of flat-headed vehicles identified in

600000 this paper, which indicate that the research results of this
paper are not limited by region and time and have universal
400000
applicability.
200000 With reference to the relationship between the above
variables and the HIC value, the nonlinear relationship
0 between the HIC and vehicle speed was fitted using multiple
1 2 3 regression analysis for different vehicle types and collision
Vehicle type angles, as shown in Table 11. According to the determination
Figure 9: Relationship between vehicle model and the HIC. coefficient R2 between different expressions (the greater the
value is, the better the goodness of fit), the optimal rela-
tionship models of the HIC, vehicle speed, and collision
angle corresponding to different models are determined as
collision position and pedestrian speed, which has little follows.
impact on pedestrian injury. Therefore, it will be eliminated
in the later analysis. (1) Short head type:
Figures 8–10 show the relationship between significant θ
risk factors and the HIC. Figure 8 shows that the HIC value HIC  −1105.442 + 105.627VS − 3.196V2S + 0.032V3S 
90∘
gradually increases with increasing vehicle speed, and the
θ
increasing trend is increasingly larger. As shown in Figure 9, + 1 − 0.006V3.276 .
a flat head, high head, and short head are more likely to cause 90∘ S

injury to the pedestrian head, while a high head and short (7)
head have the same impact on the pedestrian head. As shown
in Figure 10, pedestrian injury is more serious when the (2) High head type:
collision angle is 0° than when the collision angle is 90°; that θ
0.0001VS 
3.831
is, injury caused by a rear collision between the vehicle and HIC 
90∘
the pedestrian is greater than that caused by a side collision
θ
between the vehicle and pedestrian. The analysis shows that a + 1 − 4313.065 − 271.598VS + 4.424VS .
2
rear collision between cars and pedestrians easily causes 90∘
pedestrians to recline, which increases the probability of a (8)
serious collision between the head and automobile engine
cover, front windshield, or A-pillar. (3) Flat head type:
10 Mathematical Problems in Engineering

Table 10: Comparison of the conclusions in this paper and other research.
Literature Method Significant risk factor Data source
This Multiple regression analysis Vehicle speed, vehicle type (flat-headed
PC-Crash simulation
paper and correlation analysis vehicles), and collision angle (rear-end collision)
Pedestrian-vehicle collision data from
[18] Multilevel model Pedestrian age and vehicle type (larger vehicles)
2015 to 2017 in South Korea
Partial least squares Pedestrian age, pedestrian gender, vehicle type Fatal pedestrian crash in Tabriz, Iran,
[21]
discriminant analysis (light machinery), and accident time during 2014 to 2015
Vehicle speed (in excess of 40 km/h) and Pedestrian-vehicle accident cases in
[22] Multivariate logistic regression
collision location Chongqing, China, from 2011 to 2015
Vehicle type (trucks, light passenger vehicles) Fatal pedestrian accidents of low-speed
[23] Statistical tests
and light level vehicles in Japan from 2005 to 2014

Table 11: Model fitting results.

Number Model Collision angle Model expression R2
1 HIC � 5.195e0.08VS 0.958
2 HIC � −1105.442 + 105.627VS − 3.196V2S + 0.032V3S 0.980
θ � 90°
3 HIC � 8.968e− 5 V3.935
S 0.945
4 HIC � 5.616e0.069VS 0.941
Short head type −1
5 HIC � e10.922− 143.901VS 0.931
3.09
6 HIC � 0.013VS 0.905
θ � 0°
7 HIC � 0.006V3.276
S 0.935
8 HIC � 0.002V3.552
S 0.918
9 HIC � 7.222e0.071VS 0.966
0.078VS
10 HIC � 3.344e 0.975
θ � 90°
11 HIC � 0.0001V3.831
S 0.988
12 HIC � 4.43e0.075VS 0.969
High head type
13 HIC � 4313.065 − 271.598VS + 4.424V2S 0.934
14 HIC � −72412.624 + 5232.291VS − 107.569V2S 0.67V3S 0.920
θ � 0°
15 HIC � 0.002V3.622
S 0.909
16 HIC � 0.002V3.608
S 0.924
17 HIC � 39206.161 − 1867.829VS + 20.139V2S + 0.119V3S 0.963
18 HIC � 57651.927 − 3377.693VS + 55.218V2S − 0.152V3S 0.982
θ � 90°
19 HIC � 17718.827 − 449.132VS − 3.278V2S + 0.155V3S 0.984
20 HIC � 56592.336 − 2915.804VS + 40.36V2S − 0.079V3S 0.978
Flat head type
21 HIC � −458713.785 + 34542.659VS − 735.559V2S + 4.804V3S 0.966
22 HIC � 2946.727e0.047VS 0.910
θ � 0°
23 HIC � −729141.994 + 54160.807VS − 1112.044V2S + 7.001V3S 0.907
24 HIC � −871724.682 + 63661.079VS − 1297.743V2S + 8.087V3S 0.890

θ 2 3
HIC � 􏼐17718.827 − 449.132VS − 3.278VS + 0.155VS 􏼑
90∘
(9)
θ
+ 􏼠1 − ∘ 􏼡􏼐−458713.785 + 34542.659VS − 735.559V2S + 4.804V3S 􏼑.
90

5.2. Accident Risk Assessment. According to the data ob- function of the HIC ordered samples, and draws the vari-
tained in the tests, combined with the HIC security threshold ation curve of the minimum error function with the clas-
introduced in Section 3.1, the data with HIC ≤ 1 were sification number k, as shown in Figure 11.
screened out and classified as a separate group, and the risk As shown in Figure 11, when k � 3 and 4, the curvature of
level was determined as level I. The remaining 183 groups of the minimum error function curve changes significantly. To
data were numbered according to the HIC value from small further determine the optimal classification number, the
to large, and the HIC ordered sample was generated, which minimum error function ratio β (k) between adjacent
was denoted as xi (i � 1, 2, . . ., 183). classification numbers is calculated as shown in Table 12.
According to the specific implementation steps of According to Table 12, β (4) is greater than β (3), so k � 4
Fisher’s optimal segmentation algorithm introduced in is the optimal classification number. In addition, the clas-
Section 3.2, this study uses MATLAB to write the algorithm sification numbers corresponding to the optimal classifi-
code, calculates the class diameter and minimum error cation of ordered samples can be obtained from Table 6, and
Mathematical Problems in Engineering 11

140

Minimum error function

120
100
80
60
40
20
0
2 3 4 5 6 7 8 9
Classification number k
Figure 11: Relationship between the minimum error function and the classification number.

Table 12: Classification results.

k Minimum error function Classification β value
2 132.125 {1∼117} {118∼183} -
3 52.271 {1∼117} {118∼171} {172∼183} 1.69
4 30.856 {1∼64} {65∼117} {118∼171} {172∼183} 1.77
5 17.451 {1∼64} {65∼94} {95∼117} {118∼171} {172∼183} -

Table 13: Risk evaluation standards for vehicle-pedestrian 10%. Considering that the error rate cannot accurately
accidents. measure the degree of error classification of accident risk
Accident risk level HTC threshold Pedestrian injury level using risk assessment methods, this paper proposes a new
Class I ≤1000 Mild or moderate
index to evaluate the accuracy of accident risk assessment
Class II (1000, 5216] Heavier methods, namely, the error classification degree α, as shown
Class III (5216, 20084] Severe in formula (10). According to formula (10), the degree of
Class IV (20084, 144763] Critical error classification is 2.59%, which is within the acceptable
Class V >144763 Extreme range, thus verifying the accuracy of the risk assessment
method of automobile and pedestrian accidents proposed in
this paper.
then the corresponding HIC index threshold can be de-
termined. Based on the above research results in combi- ni1 ηij FNij
α i ≠ j, (10)
nation with the AIS injury classification standard (as shown TN + ni1 ηij FNij
in Table 1), the automobile and pedestrian accident risk
classification evaluation method is presented, as shown in where TN is the number of correctly graded cases; FNij is the
Table 13. number of cases of the i pedestrian injury level misclassified
at the j risk level; n is the number of risk levels divided; and
6. Case Analysis ηij is the corresponding weight coefficient; the calculation
formula is
To evaluate the effectiveness of the accident risk classifica- |j − i|
tion evaluation method proposed in this paper, 50 typical ηij  . (11)
n
vehicle-pedestrian accidents are selected for verification (as
shown in Table 14), including 20 short-head automobile The research results of this paper can not only provide a
accidents, 19 high-head automobile accidents, and 11 flat- scientific judgment basis for accident analysis and treatment
head automobile accidents. First, according to the infor- but also provide a reference for speed limit management. For
mation of each accident case, combined with models (7)–(9) example, in the absence of effective accident scene evidence
fitted in Section 5.1, the HIC value corresponding to each and video surveillance, a judicial appraisal institution can
accident is calculated, and then the accident risk level is determine the corresponding HIC range from Table 13
determined according to Table 13. Finally, the evaluation according to the pedestrian injury level in the accident.
method of automobile pedestrian accident risk classification Then, combined with the car accident and collision shape,
is verified using the actual injury level of pedestrians. Among using the optimal fitting model, this research deduced the
them, the actual injury level of pedestrians is evaluated using range of speed in a trouble-causing car accident, and the
the trauma scoring system Version 3.0 according to the calculation result is helpful to a traffic police department
injury report issued by a hospital or medical examiner. responsible for both sides of the accident. Similarly, for
It can be seen from the evaluation results in Table 14 that sections with a high incidence of vehicle-pedestrian acci-
the risk level corresponding to 5 accidents calculated using dents, the maximum allowable speed that can guarantee
the accident risk evaluation method proposed in this paper is HIC≤1000 (the accident risk is grade I) can also be deduced
not consistent with the actual pedestrian injury level, based on the optimal model for different models to for-
namely, cases 22, 30, 39, 43, and 46, with an error rate of mulate the speed management scheme.
12 Mathematical Problems in Engineering

Table 14: Case validation.

Number Accident speed (km/h) Collision form Model Actual pedestrian injury level HIC Accident risk level
1 75 Rear-end collision Short head type Severe 8334 III
2 84 Side collision Short head type Heavier 4183 II
3 50 Rear-end collision Short head type Heavier 2208 II
4 47 Side collision Short head type Mild 121 I
5 56 Side collision Short head type Moderate 407 I
6 70 Side collision Short head type Heavier 1604 II
7 82 Side collision Short head type Heavier 3710 II
8 95 Side collision Short head type Severe 7521 III
9 34 Rear-end collision Short head type Moderate 624 I
10 51 Rear-end collision Short head type Heavier 2356 II
11 49 Rear-end collision Short head type Heavier 2066 II
12 80 Rear-end collision Short head type Severe 10296 III
13 85 Rear-end collision Short head type Severe 12558 III
14 44 Side collision Short head type Mild 81 I
15 48 Side collision Short head type Mild 140 I
16 66 Side collision Short head type Heavier 1144 II
17 71 Rear-end collision Short head type Severe 6964 III
18 77 Rear-end collision Short head type Severe 9084 III
19 89 Side collision Short head type Severe 5539 III
20 40 Side collision Short head type Mild 54 I
21 37 Rear-end collision High head type Mild 320 I
22 69 Rear-end collision High head type Heavier 6635 III
23 76 Rear-end collision High head type Severe 9225 III
24 99 Rear-end collision High head type Critical 20784 IV
25 105 Rear-end collision High head type Critical 24570 IV
26 73 Rear-end collision High head type Severe 8062 III
27 62 Rear-end collision High head type Heavier 4480 II
28 55 Side collision High head type Moderate 465 I
29 78 Rear-end collision High head type Severe 10044 III
30 74 Side collision High head type Severe 1449 II
31 55 Side collision High head type Moderate 465 I
32 53 Side collision High head type Moderate 403 I
33 30 Side collision High head type Mild 46 I
34 50 Side collision High head type Moderate 323 I
35 47 Rear-end collision High head type Heavier 1321 II
36 49 Rear-end collision High head type Heavier 1627 II
37 51 Side collision High head type Moderate 348 I
38 63 Side collision High head type Moderate 782 I
39 65 Side collision High head type Heavier 882 I
40 76 Rear-end collision Flat head type Critical 26780 IV
41 72 Side collision Flat head type Extreme 186242 V
42 58 Side collision Flat head type Extreme 170884 V
43 48 Side collision Flat head type Severe 165750 V
44 69 Side collision Flat head type Extreme 182041 V
45 78 Side collision Flat head type Extreme 196299 V
46 34 Rear-end collision Flat head type Severe 54247 IV
47 83 Rear-end collision Flat head type Critical 87926 IV
48 88 Rear-end collision Flat head type Extreme 158663 V
49 97 Rear-end collision Flat head type Extreme 355531 V
50 77 Rear-end collision Flat head type Critical 33126 IV

In addition, because the collision between an automobile protecting pedestrians. For example, by capturing the
hood and pedestrians easily causes fatal injuries, the space movement characteristics of pedestrians and vehicles,
between the hood and various parts in the engine com- computer vision technology can be used to predict potential
partment can be considered in the design of a vehicle to accidents between pedestrians and vehicles at signalized
ensure that the hood has enough deformation space for intersections, which can be used for the development of
cushioning the impact force. Of course, improving active collision warning systems in connected vehicle environ-
safety technology and road design will also play a role in ments [28].
Mathematical Problems in Engineering 13

7. Conclusion Conference, pp. 725–751, SAE Warrendale PA, California,

USA, November 1975.
Based on accident case investigations, this research uses PC- [3] H. Cheng and A. L. Rizer, Articulated Total Body Model
Crash to carry out vehicle-pedestrian collision simulation Version V User’s Manual, pp. 637–648, United States Air
tests under different accident patterns, revealing the sig- Force Research Laboratory, USA, 1998.
nificant risk factors (i.e., vehicle speed, collision angle, and [4] R. R. Mchenry, “Analysis of the dynamics of automobile
vehicle type) affecting pedestrian injury. For different vehicle passenger restraint systems,” in Proceedings of the 7th Stapp
Car Crash Conference, pp. 420–431, California, USA, 1963.
types, a multivariate relationship model between the HIC,
[5] H. Appel, G. Sturtz, and L. Gotzen, “Influence of impact speed
vehicle speed, and collision angle is built, and a classification and vehicle parameter on injuries of children and adults in
evaluation method of vehicle-person accident risk based on pedestrian accidents,” in Proceedings of the 2nd IRCOBI
the HIC is proposed and verified. Research results of this Conference, pp. 83–100, Hong Kong, China, September 1975.
paper can promote pedestrian injury reduction in accidents [6] D. Severy and H. Brink, “Auto-pedestrian collision experi-
and provide guidance on speed limit measures. ments,” SAE Transactions, vol. 75, pp. 323–371, 1967.
Although this research has made some innovative [7] Y. Matsui, H. Ishikawa, and A. Sasaki, “Proposal of injury risk
achievements, there are still the following deficiencies. For curves for evaluating pedestrian femur/pelvis injury risk using
example, this paper only selects Chinese 90th% adult men as EEVC upper legform impactor based on accident recon-
representatives for the crash tests, without considering adult struction,” International Journal of Crashworthiness, vol. 11,
women, children, the elderly, and other groups. As a result, no. 2, pp. 97–104, 2006.
[8] R. Cheng, Y. Pan, and T. Wang, “A design method for the
the factors involved in the proposed accident risk evaluation
roadside clear zone based on accident simulation analysis,”
method are not sufficiently comprehensive and ignore
Mathematical Problems in Engineering, vol. 202116 pages,
factors of pedestrian gender or pedestrian age that may affect Article ID 2605095, 2021.
the accident risk. In addition, the values of the variables were [9] G. Cheng, R. Cheng, Y. Pei, and L. Xu, “Probability of
relatively low in the test process, and there was no relevant roadside accidents for curved sections on highways,” Math-
research on occupant movement. Therefore, follow-up re- ematical Problems in Engineering, vol. 2020, Article ID
search work can consider increasing the value range of 9656434, 18 pages, 2020.
variables and combining PC-Crash and MADYMO software [10] J. Mandelı́k and M. Bundzel, “Application of neural network
to conduct in-depth research on the injury mechanisms of in order to recognise individuality of course of vehicle and
different pedestrian groups and the response of drivers in the pedestrian body contacts during accidents,” International
car before and after the accident. Journal of Crashworthiness, vol. 24, no. 2, pp. 221–234, 2019.
[11] S. Xu, X. Jin, C. Qin, and X. Chai, “Personalized customization
method of hybrid human model for pedestrian-vehicle ac-
Data Availability cident reconstruction,” Journal of Mechanics in Medicine and
Biology, vol. 21, no. 02, Article ID 2150009, 2021.
The data used to support the findings of this study are [12] D. Grindle, W. Pak, B. Guleyupoglu et al., “A detailed finite
available from the corresponding author upon request. element model of a mid-sized male for the investigation of
traffic pedestrian accidents,” Proceedings of the Institution of
Conflicts of Interest Mechanical Engineers - Part H: Journal of Engineering in
Medicine, vol. 235, no. 3, pp. 300–313, 2021.
The authors declare that they have no conflicts of interest. [13] P. Panday, A. Vikram, A. Chawla, and S. Mukherjee, “Pre-
diction of lower extremity injuries in car-pedestrian crashes
Acknowledgments – real-world accident study,” Traffic Injury Prevention, vol. 22,
no. 2, pp. 173–176, 2021.
This study was supported by the National Nature Science [14] M. Lalwala, A. Chawla, P. Thomas, and S. Mukherjee, “Finite
Foundation of Guangxi Province (Nos. element reconstruction of real-world pedestrian accidents
2022GXNSFBA035640, 2019GXNSFBA245050, and using thums pedestrian model,” International Journal of
2019JJB160047), the Guangxi University Young and Middle- Crashworthiness, vol. 25, no. 4, pp. 360–375, 2020.
[15] S. Zhu, “Analyse vehicle–pedestrian crash severity at inter-
aged Teachers Scientific Research Basic Ability Improve-
section with data mining techniques,” International Journal of
ment Project (2022KY0193), the Guangxi Science and Crashworthiness, vol. 5, no. 26, pp. 1–9, 2021.
Technology Base and Talent Special Project (Nos. [16] A. Sheykhfard, F. Haghighi, T. Nordfjærn, and
AD20159035 and AD22035153), and the Fund for Less- M. Soltaninejad, “Structural equation modelling of potential
developed Regions of the National Natural Science Foun- risk factors for pedestrian accidents in rural and urban roads,”
dation of China (No. 71861006). International Journal of Injury Control and Safety Promotion,
vol. 28, no. 1, pp. 46–57, 2021.
References [17] N. Kamboozia, M. Ameri, and S. M. Hosseinian, “Statistical
analysis and accident prediction models leading to pedestrian
[1] Traffic Management Bureau of the Ministry of Public Security, injuries and deaths on rural roads in Iran,” International
Road Traffic Accident Statistical Annual Report, OECD, Journal of Injury Control and Safety Promotion, vol. 27, no. 4,
Beijing, China, 2021. pp. 493–509, 2020.
[2] H. Pritz, C. Hassler, J. Herridge, and E. Weis, “Experimental [18] S. Park and D. Ko, “A multilevel model approach for in-
study of pedestrian injury minimization through vehicle vestigating individual accident characteristics and neighbor-
design,” in Proceedings of the 19 Th Stapp Car Crash hood environment characteristics affecting pedestrian-vehicle
14 Mathematical Problems in Engineering

crashes,” International Journal of Environmental Research and

Public Health, vol. 17, no. 9, p. 3107, 2020.
[19] D. Lasota, A. Al-Wathinani, P. Krajewski, K. Goniewicz, and
W. Pawlowski, “Alcohol and road accidents involving pe-
destrians as unprotected road users,” International Journal of
Environmental Research and Public Health, vol. 17, no. 23,
pp. 8995–95, 2020.
[20] J. Hasani, S. Erfanpoor, A. Rajabi et al., “Spatial analysis of
mortality rate of pedestrian accidents in Iran during
2012–2013,” Traffic Injury Prevention, vol. 20, no. 6,
pp. 636–640, 2019.
[21] M. Jamali-Dolatabad, H. Sadeghi-Bazargani, and
P. Sarbakhsh, “Predictors of fatal outcomes in pedestrian
accidents in Tabriz Metropolis of Iran: application of pls-da
method,” Traffic Injury Prevention, vol. 20, no. 8, pp. 873–879,
2019.
[22] C. Feng, A. Duan, J. Qiu et al., “Investigation of head injuries
suffered by pedestrians in traffic accidents via videos,” In-
ternational Journal of Crashworthiness, vol. 25, no. 6,
pp. 680–688, 2020.
[23] Y. Matsui and S. Oikawa, “Situational characteristics of fatal
pedestrian accidents involving vehicles traveling at low speeds
in Japan,” Traffic Injury Prevention, vol. 20, no. sup1, pp. 1–6,
2019.
[24] Editorial Department of China Journal of Highway and
Transport, “Review on China’s automotive engineering re-
search process: 2017,” China Journal of Highway and
Transport, vol. 30, no. 6, pp. 1–197, 2017.
[25] G. Cheng, R. Cheng, Y. Pei, L. Xu, and W. Qi, “Severity
assessment of accidents involving roadside trees based on
occupant injury analysis,” PLoS One, vol. 15, no. 4, Article ID
e0231030, 2020.
[26] G. Cheng, R. Cheng, L. Xu, and W. Zhang, “Risk assessment of
roadside accidents based on occupant injuries analysis,”
Journal of Jilin University (Engineering and Technology Edi-
tion), vol. 51, no. 3, pp. 875–885, 2021.
[27] D. Liu, Y. Li, H. Zhao, and X. Zhu, “Relativity analysis of
pedestrian head injuries and the headform to bonnet top test
method,” Journal of Highway and Transportation Research
and Development, vol. 2004, no. 1, pp. 98–101, 2004.
[28] S. Zhang, M. Abdel-Aty, Y. Wu, and O. Zheng, “Modeling
pedestrians’ near-accident events at signalized intersections
using gated recurrent unit (GRU),” Accident Analysis &
Prevention, vol. 148, Article ID 105844, 2020.