2014

2014
Eighth
Eighth
International
International
Conference
Conference
on Next
on Next
Generation
Generation
Mobile
Mobile
Applications,
Apps, Services
Services
and and
Technologies
Technologies

Detection of Vulnerable Road Users in Smart Cities

Francisco Guayante, Arnoldo Díaz-Ramírez Pedro Mejía-Alvarez

Department of Computer Systems Department of Computer Sciences
Instituto Tecnológico de Mexicali CINVESTAV-IPN
Av. Tecnológico s/n Col. Elías Calles Av. Instituto Politécnico Nacional 2508, Col. Zacatenco
Mexicali, B.C., México 21376 México, D.F. 07360
{fguayante, adiaz}@itmexicali.edu.mx [email protected]

Abstract—Starting from 2008, more than half of the world’s of the information and communication technology (ICT) to
population now lives in urban areas, and this number is expected make a city smarter. A smart city is defined by Washburn
to grow for the next decades. To the extent that the population et al. as “the use of smart computing technologies to make
of a city grows, new problems arise, which include scarcity of
resources, pollution, and traffic congestion. One of the most the critical infrastructure components and services of a city
important problems of big cities are road traffic injuries, which –which include city administration, education, healthcare,
is the eighth leading cause of death globally, and the main cause public safety, real estate, transportation, and utilities– more
of death for young people, mainly in middle and low income intelligent, interconnected, and efficient” [13]. Smart comput-
countries. Vulnerable road users (VRUs) are among the users ing is referred as “a new generation of integrated hardware,
at higher risks of traffic accidents. In order to cope with the
problems of the growing urban communities, the concept of software, and network technologies that provide IT systems
smart cities has emerged. A smart city is based on the use of with real-time awareness of the real world and advanced
smart computing technologies, such as Intelligent Transportation analytics to help people make more intelligent decisions about
Systems and Vehicular Ad hoc Networks. In this paper, we alternatives and actions that will optimize business processes
propose a model to be used in smart cities, to detect if a VRU and business balance sheet results” [13].
intends to cross a road in a risky zone, and to issue alerts to the
vehicles nearby. The proposed model is cost effective, and is able In this paper, we propose a model to detect if a VRU intends
to detect a VRU at risk in a short period of time. The evaluation to cross a road in a risky zone, and to issue alerts directed
of the proposed model shows that it performs correctly. to the vehicles nearby. The proposed model was developed
Index Terms—Smart Cities, VRUs detection, VANETs in the context of smart cities. It is based on the use of two
emerging technologies: the Intelligent Transportation Systems
(ITS) paradigm [8], which consists in the use of information
I. I NTRODUCTION and communication-based solutions to improve the interac-
Nowadays, more than half of the world’s population lives tion between vehicles and road users; and the Vehicular Ad
in urban areas, and the migration from rural zones to cities is hoc Networks (VANETs) technology [3]. VANETs comprise
expected to continue for the next decades [1]. However, bigger vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I)
cities generate new kinds of problems, such as scarcity of communication based on wireless local area network technolo-
resources, pollution, health, traffic congestion and inadequate gies.
infrastructure, to mention a few. This paper is organized as follows: Section I presents
One of the most important problems of big cities are road an overview of the related work. Section II introduces the
traffic injuries, which is the eighth leading cause of death proposed system. In Section III, the results of its evaluation are
globally, and the main cause of death for young people. Every shown. Finally, the conclusions and future work are discussed
year, more than 1.2 million people die on the world’s roads, in Section V.
and 50 million people are injured. Middle-income countries
have 52% of the world’s vehicles, and 80% of the world’s II. R ELATED W ORK
road traffic deaths [2]. Also, there are different road user In the past few years, some proposals have been published
groups, with differences in fatality rates among them. The to protect pedestrians using systems based on information
users at greatest risk are the vulnerable road users (VRU’s), technology solutions. These proposals may fall into one of
such as pedestrian and cyclists. In middle and low income the three categories proposed by Carsten in [4]: in-vehicle sys-
countries, the rate of VRUs deaths is greater, in part because tems, carried-by-the-pedestrian systems, and indirect systems.
less educated people disobey traffic signals and cross streets
through dangerous zones. A. In-vehicle systems
To face the problems generated by the rapid urban popu- In-vehicle systems are constituted of sets of sensors or
lation growth, recently we have seen an interest in the use cameras, and processing units mounted within the vehicles, for

978-1-4799-5073-7/14 $31.00 © 2014 IEEE 307

DOI 10.1109/NGMAST.2014.60
the purpose of detecting VRUs. These systems alert drivers systems connected to the road infrastructure.
when risky situations are detected, and suggest or perform Meissner et al. in [10] proposed a method to prevent
evasive actions. One drawback of these kind of systems is the pedestrian accidents at intersections. To recognize and track
short range of detection of the sensors within the vehicle, as moving objects, a network of laser scanners is deployed several
well as the signal’s distortion generated at high speeds. Also, meters above the ground, allowing an aerial view of the area of
some measurements may be incomplete due to the presence interest. The network is able to generate a three-dimensional
of other vehicles, which may interfere with the detection of profile of the scene, regardless of the light conditions. The
the VRUs. method is able to distinguish dynamic objects (pedestrians)
In [6], Fuerstenberg et al. introduced a pedestrian detection from static objects. The pedestrians are identified by means of
and classification system. Their proposal uses laser scanner density and dimensions of the points detected by the sensors.
sensors, which allows a 180 degrees horizontal field of view. Once the system has detected a pedestrian, the information
Based on the images taken by the laser scanners, several regarding his movement and position is transmitted to the vehi-
pedestrian recognition algorithms are able to process signals cle using V2I communication. The sensors and the algorithms
from objects moving at high speeds, even over long distances. used allow the detection and tracking of multiple pedestrians.
The pedestrians movements may be predicted to determine In [9], Hugues et al. proposed a detection system, where
whether or not they are at risk. microwave detectors are placed in traffic signals to identify
In [7], Garate et al. proposed a support system for drivers pedestrians at the time they approach to a crossing. The
using sensor fusion techniques, to process the data collected information generated by the sensors was used to make time
by a short-range radar, and the monochrome images captured alterations in traffic signals for vehicles, in order to improve
by a camera. The radar system is placed in the center of the the safety of pedestrians. The results showed that this is a
front of the vehicle, and has a single beam with a range of viable method because the microwave sensor can detect a
40 meters, and a horizontal view of 50 degrees. It operates at moving object in a way that the system is capable to take
a frequency of 100 Hz, and it is capable to detect up to 32 fast decisions.
objects. Using the information obtained by the radar (distance
and angle position of the vehicle), and the vehicle speed, it is III. P ROPOSED S YSTEM
possible to determine if the detected object is in danger of a As mentioned previously, our proposal is based on the
collision. Intelligent Transportation Systems (ITS) paradigm, and on the
Vehicular Ad hoc Networks (VANETs) technology. VANETs
B. Carried-by-the-pedestrian systems communication is based on the IEEE 802.11p protocol, which
To detect pedestrians and prevent them from any road- is an amendment to the IEEE 802.11 standard to add wireless
related risk, some components of these kind of systems must access in vehicular environments (WAVE). In a VANET,
be carried by the VRUs. The proliferation of smartphones and vehicles communicate with other vehicles using the vehicle-
similar mobile devices allows the implementation of carried- to-vehicle (V2V) model. Also, vehicles communicate with
by-the-pedestrian systems. However, the system may not work fixed equipment next to the road, referred to as road side unit
if the pedestrian does not carry the device. (RSU), using the vehicle-to-infrastructure model (V2I). The
In [5], Fardi et al., proposed a cooperative sensing system main system components of the V2I model are the application
composed of an infrared camera placed inside the vehicle, and unit (AU), the on board unit (OBU), and the roadside unit
Wireless Personal Area (WPAN) modules, to determine the (RSU). Commonly the RSU hosts an application that provides
relative position of the VRU. The vehicle and the VRUs are some services that are used by the AU, which is a device
equipped with these modules, respectively. The data collected equipped within the vehicle. The AU uses the communication
by the sensors are processed using data fusion methods, with capabilities of the OBU, which is a WAVE device mounted
the objective of achieving a reliable real-time detection of on the vehicle, aimed for exchanging information with RSUs
VRUs around the vehicle. As a consequence of the detection or other OBUs [3].
of a VRU at risk, a warning message is sent, directed to both According to Sherborne [11], a VRU detection system
the driver and the VRUs. based on the indirect system model, should consider different
A Collaborative Context Predictor was introduced by Voigt- scenarios where the vehicle and the pedestrian interact. In
mann et al. in [12]. They proposed a proactive approach to particular, a pedestrian detection system at crossings points
detect pedestrians who are dangerously close to the street, and should be able to:
try to predict the pedestrian’s next step using her/his context • Identify pedestrians approaching to the crossing point.
information (movement and orientation). The pedestrians must • Identify pedestrians waiting to cross the road.
carry a smartphone in their left trouser pocket, to gather • Obtain a measure of how many pedestrians intend to cross
movement data. The objective of the proposed system is to the road.
detect a risk situation in a very short period of time, to avoid • Identify the situations when pedestrians no longer wish to
a collision. cross the road.
• Identify the situations when all pedestrians have crossed
C. Indirect systems the road.
These systems are aimed to improve the interaction between Based on these observations, we propose a VRUs detection
the VRUs and the vehicles, through the use of detection and monitoring system for smart cities, to be used at risky

308
Figure 1. System architecture Figure 2. Detection system using four non-contact temperature sensors

or prohibited crossing areas. This system has been developed In order to select the most suitable sensors to successfully
using the following design guidelines: detect VRUs, we evaluated different types of sensors. Some
• It is based on the indirect system model. This ensures of the them were the Passive InfraRed (PIR) sensors, which
that the VRU is detected without the need to carry any device measure the infrared light emanating from objects. They are
with him. In contrast to the in-vehicle system model, indirect cheap sensors that detect the presence of heat from an object or
systems offer a better perspective of the situation, and are able body nearby. They are also capable of detecting the movement
to alert vehicles long before they reach the risk zone. of people when a temperature change occurs. Motion detectors
• It is based on the V2I architecture. The proposed system usually use PIR sensors. However, the use of PIR sensors does
sends alerts to vehicles when a VRU at risk is detected, using not allow the detection of multiple pedestrians in the sensed
the VANET’s technology. area. Additionally, they are not able to detect a pedestrian if
• The system is affordable. Many of the already proposed he is not moving, and if the body temperature is similar to the
systems are attractive and efficient. However, they are based ambient temperature, the pedestrian would not be detected.
on the use of very expensive sensors. One key design concern In contrast, non-contact temperature sensors have many of
of the proposed system is cost. the advantages of the PIR sensors, and overcome many of their
• The system has a low computational complexity. Thus, it is disadvantages. They are constituted by a set of temperature
able to detect, track and determine the intentions of pedestrians sensors distributed in a matrix. Each temperature sensor has
in a short period of time. an angle of view, and it is capable of measuring the heat
The architecture of the proposed system is shown in Fig. 1. radiation from the surface of an object. The advantage of these
As it can be observed, it is comprised of four modules, which sensors, in contrast with other infrared sensors, is that the
are described next. latter are not able to detect static objects because they only
detect drastic r temperature changes. The non-contact infrared
temperature sensors can detect an object whether or not an
A. VRU detection module object is moving.
This module is responsible of performing the VRUs de- These devices are constituted by multiple sensors, in a way
tection functions, as well as the classification functions. The that the output resembles a low resolution thermal image. This
module consists of a vision unit, two communication interface allows not only to detect motion but also to determine in which
units, and a processing submodule. direction the detected objects are moving, and to calculate
Vision unit: The vision unit must determine if one or more the distance between the object and the sensor. Additionally,
VRUs are in the detection zone. The sensors used should allow the use of several sensors may increase the detection area, as
the system, by means of the acquired data, to determine if an shown in Fig. 2.
object is in motion, if it is a VRU or not, and the motion’s In the proposed system, we assume the use of one or more
direction of the detected VRU. non-contact temperature sensors.
As discussed previously, many of the proposals that have Processing submodule: It consists of both physical and
been published to date use complex techniques (e.g. image logical components that are responsible of processing the data
processing), use expensive sensors, or may require drastic collected by the vision unit. It comprises a processor unit,
changes to the road infrastructure. In contrast, the proposed and three algorithms: a detection algorithm, a classification
model is able to alert vehicles when a VRU at risk is algorithm and a tracking algorithm.
detected, which requires the use of low complexity algorithms. The processor unit (a microprocessor or computer), is where
Additionally, the system uses low cost and highly available the algorithms are executed to process the data gathered by
sensors. Since it is based on the V2I communication model, the sensor(s). The detection algorithm determines if there is
it uses the VANET’s infrastructure. a moving object in the monitored area. The classification

309
algorithm uses the information provided by the detection
algorithm as input, compares the data with a temperature
threshold, and determines whether or not the moving object
is a pedestrian. If a VRU is detected, the tracking algorithm
follows his movement in order to determine if he is directed
towards a risk zone (the street), if he is moving away from
the danger zone, or if he is going out of the detection area.
Interface units: These units are composed of hardware and
software resources, that allow the connection between the
system’s components. For instance, an interface unit connects
the vision unit and the drive processor submodule. Also,
another interface unit connects the processor submodule to
the road side unit module.
Figure 3. Distribution of the OMRAN D6T-44L device

B. Road Side Unit (RSU) Module

The RSU module is part of the road infrastructure. It is Classification algorithm: If the detection algorithm deter-
able to process the data gathered by the sensor(s), and to send mines that there is a moving object within the detection zone,
messages directed to the vehicles. Once the detection module then the classification algorithm checks if the temperature of
has generated a warning, it is sent to the RSU module, which the moving object is within the temperature range characteris-
in turn will send the alert message to the vehicles nearby. tic of a human being. If so, the information is sent to the next
The vehicles that receive a warning message will re-transmit process.
it to other vehicles using the V2V model. These messages are Tracking algorithm: Once the system has detected the
received by the vehicle’s OBUs. presence of a VRU, the tracking algorithm checks element
by element of the array, compares previous measurements with
current ones, to determine in which direction the VRU moves.
C. On Board Unit (OBU) Module An example of the detection zone is shown in Fig. 5.

It is the unit installed within the vehicle, which is respon- IV. E VALUATION
sible for receiving information from the RSUs, or from other
OBUs. The OBU module is composed of a transmitter/receiver In order to evaluate the proposed model, we built a proto-
unit, a submodule processor, and several submodules or actu- type, which is discussed next.
ator units. In this case, when the system detects a VRU trying
to cross the street, the OBU will receive the alert message, A. Vision Unit
and will send it to the AU, which will inform the driver about We use an OMRON D6T-44L device as a vision unit, which
the risk situation. is composed of an array of 16 thermal sensors distributed in
a 4x4 matrix, as shown in Fig. 3. This particular device has
a horizontal viewing angle of 44.2 °, and a vertical viewing
Proposed algorithms angle of 45.7 °.
As already mentioned, within the processor submodule there The device is able to detect an object up to a distance of
are three logical units that allow the system to perform its 10 meters. However, the detected temperatures vary as the
functions properly. This strongly depends on the proposed distance increases. At a distance of 1 cm, the measured surface
algorithms within the submodule. Regarding the algorithms, it temperature of the human body is approximately 32 °C. On the
is important to point out that they were designed and developed other hand, at a distance of 5 meters, the measured temperature
assuming the use of an array of non-contact temperature is of 28 °C. For these reasons, the motion detection threshold
sensors. and human temperature threshold should consider the distance
Master algorithm: It is in charge of the decision taking between the sensor and the surface of the detected object. The
process, once the data have been collected by the vision unit. D6T-44L device detects objects whose surface temperature is
In addition, it is responsible for storing the data measured by between 5 °C and 50 °C, and withstands temperatures between
the sensor array. -10 °C and 60 °C.
Detection algorithm: This algorithm compares the current The communication interface that the sensor uses to com-
measurements captured by the system, with the prior cycle municate with other devices or control units is the Inter-
measurements for each sensor array element. If the difference Integrated Circuit (I2C), which is a low speed communication
between the current measurement and the previous one is bus developed by Phillips. This interface uses only two com-
located within a defined detection threshold, the algorithm munication wires: the Serial Clock Line (SCL), and the Serial
determines that there is a moving object within the detection Data Line (SDA). The SCL is responsible for controlling data
zone. In that case, the classification algorithm is invoked. transmission through a clock signal, which determines the start

310
and the end of a transmission. The I2C channel works using • The distance between the sensor and the ground was set
the Master and Slave scheme. The master device is responsible to of 2.35 meters, which provided a sensing area of 1.2
over the SCL line. Up to 255 slave devices can be connected square meters.
to the same bus, as long as there is a master device. • With the purpose of the evaluation, the simplest scenario
In the prototype, we used a personal (notebook) computer was considered: the detection of a single pedestrian mov-
as the processing unit. The communication interface between ing in one direction. To detect more than one pedestrian,
the computer, and the I2C communication bus of the D6T- a multi-threaded application may be constructed, where
44L device, was implemented using the Arduino platform. each thread will be in charge of tracking the movements
The Arduino programing language includes a library, named of one pedestrian.
Wire, that simplifies the communication through the I2C bus. • To evaluate the system, we assumed that the movement
One limitation of the Wire library is that it only supports a of the VRU always started in any of three points: the
maximum of 32 bytes over the bus, whereas the D6T-44L opposite side of the street, and both lateral sides. An
device transmits 35 bytes for each read cycle. To solve this object coming from the street was not considered of
problem, a library that allows extending the capabilities of interest.
Wire was used, which is called WireExt.

Figure 4. D6T-44L’s output data structure

Fig. 4 shows the scheme under which the D6T-44L com-

municates. First, a byte containing the slave’s device address
is sent. Then, a byte that indicates the instructions is sent
(this is done only once). Afterwards, it starts a cycle of data
transmission, where the first two bytes are designated for
the sensor’s reference temperature, and two bytes of data for
each element of the array. The last byte is known as PEC
Figure 5. Detection Area
byte, which is used to check if the information was received
correctly.
The motion detection threshold used in the experiments was
The data received by the sensor are converted to Celsius
set to 0.5 °C. As mentioned before, the previous measurement
degrees and stored along with the reference temperature within
and the current ones were compared. If the difference was
an array of 17 elements. The data received by the processing
greater than 0.5 °C, then the system concluded that a moving
unit are processed by the algorithms, which were developed
target was detected. To determine if the detected object is a
using Processing V2. This is a programming language and an
pedestrian, we used the temperature range for human skin
IDE based on he Java platform. The processing unit receives
(from 25.5 °C to 30 °C). To reduce the number of false
the data from the Arduino device via the serial port of the
positives, the range of human detection was modified to 26.5
computer. The reading process goes on while there are still
°C to 30 °C.
data available on the serial port.
The detection area for a 4x4 array of sensors can be seen
in Fig. 5. The point A is the closest to the street, whereas
B. Experiments
point B is the opposite side of the street. Points C and D are
To perform the experiments to evaluate the proposed system, the lateral sides of detection area. The measurement for each
the following conditions were used in a controlled environ- element was done from left to right, and from bottom to top.
ment: The first row of the array was considered Zone 1, the next one
• The average ambient temperature was 25 °C. was Zone 2, and so on. Zone 4 was assumed as the edge of
• The floor surface temperature measured by the sensor the street.
varied from 24.1 °C to 24.5 °C. That is, in normal The evaluation of the prototype was performed under the
conditions and without the presence of motion. The following scenarios (again refer to Fig. 5):
measured temperatures when there was no motion was in • Scenario 1: The pedestrian enters from B and walks
that range. If the collected temperature was outside this straight to A.
range, then we assumed that there was a moving object • Scenario 2: The pedestrian enters from B and stands still.
in the detection zone. • Scenario 3: The pedestrian enters from B and is headed
• The temperature of the human skin (wearing clothes) was towards C or D.
in the range from 25.5 °C to 28 °C, when the sensor was • Scenario 4: The pedestrian enters from B, remains mo-
place up to 4 meters away from the target. If the sensed tionless, and after a moment, he resumes his movement.
temperature fall in that range, then the system assumed • Scenario 5: The pedestrian enters from C or D and goes
that the detected object was a pedestrian. towards A.

311
 Scenario Tests Successful Tests % of Success False Positive False Negative
1 25 20 80% 5 0
2 20 20 100% 0 0
3 40 40 100% 0 0
4 60 60 100% 0 0
5 40 40 100% 0 0
6 40 40 100% 0 0
7 40 40 100% 0 0
8 60 50 83% 0 10
Total 325 310 95% 5 10
Table I
TABLE OF TESTS

• Scenario 6: The pedestrian enters from C or D and goes In this paper, we proposed a model to be used in smart
towards B. cities, to detect if a VRU intends to cross a road at a risky
• Scenario 7: The pedestrian enters from C or D and goes zone, and to issue alerts to the vehicles nearby. The model
towards the opposite side. is based on the use of Vehicular Ad hoc Networks (VANETs)
• Scenario 8: The pedestrian enters from B, C or D and technology, and on the use of non-contact temperature sensors.
returns to the same side where he started. In order to evaluate the proposed system, we built a prototype
To evaluate the performance of the system, we conducted using low cost and highly available components. The results
20 experiments for each scenario. of its showed that the model was able to detect a pedestrian at
For Scenario 1, all tests were successful. We noted that risk under controlled scenarios, and using low computational
in the cases where the VRU moves at high speed, (i.e., the complexity algorithms.
pedestrian ran), the system sometimes failed to detect him. For As future work, we intend to build a prototype that will
Scenario 2, if a pedestrian enters to the detection zone from be able to detect more than one VRU moving in different
either side, the system waits for the VRU to start moving again, directions, or moving at high speed.
storing his status and last position. For these experiments, all
the cases were detected successfully. For Scenario 3, which is
similar to Scenario 1, also all tests were successfully detected. R EFERENCES
For this scenario, we conducted forty tests: twenty when the [1] United Nations Population Fund web site. http://www.unfpa.org.
VRU was headed to C, and twenty when he was moving [2] The Global status report on road safe. World Health Organization, 2013.
[3] Saif Al-Sultan, Moath M. Al-Doori, Ali H. Al-Bayatti, and Hussien
towards D. For Scenario 4, the position of the pedestrian when Zedan. A comprehensive survey on vehicular ad hoc network. Journal
he stopped is stored, to allow the detection of the target when of Network and Computer Applications, 37:380–392, 2014.
he begins to move again. In this set of experiments, twenty test [4] O.M.J. Carsten. Can road transport informatics help vulnerable road
users? In Proceedings of the IEEE Vehicle Navigation and Information
where conducted when the user was headed towards A, twenty Systems Conference, pages 1031–1036, Oct. 1991.
when he moved towards C, and twenty when he directed [5] Basel Fardi, Ulrich Neubert, Normen Giesecke, Holger Lietz, and Gerd
towards D. Scenarios 5 and 6 are similar, and all their cases Wanielik. A fusion concept of video and communication data for vru
recognition. In Proceedings of the 11th IEEE International Conference
were successfully detected, as well as in the experiments for on Information Fusion, pages 1–8, July 2008.
Scenario 7. However, for Scenario 8, the system failed in [6] K. Ch. Fuerstenberg, K. C. J. Dietmayer, and Willhoeft V. Pedestrian
detecting 50% of the cases in which the VRU entered and recognition in urban traffic using a vehicle based multilayer laser
scanner. In Proceedings of the IEEE Intelligent Vehicle Symposium,
returned from the same side. This can be explained by the fact pages 31–35, June 2002.
that sometimes the VRU approached too much to the point at [7] Virginia Ruiz Garate, Roy Bours, and Kajetan Kietlinski. Numerical
which it was considered that he had the intention to cross the modeling of ada system for vulnerable road users protection based on
radar and vision sensing. In Proceedings of the IEEE Intelligent Vehicles
street. In the case for which the VRU is coming and returning Symposium, pages 1150–1155, June 2012.
from C or D, all tests were successful. In Table I we can [8] Sarah Hasan and Edwin Yaz. Intelligent Transportation Systems. PhD
observe a summary of results obtained after conducting the thesis, University of Arkansas, 1999.
[9] Ronald Hughes, Herman Huang, Charles Zegeer, and Michael Cynecki.
experiments. Evaluation of automated pedestrian detection at signalized intersections.
Research Report, Department of Transportation, 2001.
[10] Daniel Meissner, Stephan Reuter, and Klaus Dietmayer. Real-time
V. C ONCLUSIONS AND F UTURE W ORK detection and tracking of pedestrians at intersections using a network
of laserscanners. In Proceedings of the IEEE Intelligent Vehicles
The rapid urban growth occurred in recent years has led Symposium, pages 630–635, June 2012.
to the emergence of new problems. For instance, road traffic [11] D. J. Sherborne. Existing techniques for detecting vulnerable road users.
Drive II Project V2005, 1992.
injuries is the eighth leading cause of death globally, and the [12] Christian Voigtmann, Sian Lun Lau, and Klaus David. Evaluation of
main cause of death for young people. The vulnerable road a collaborative-based filter technique to proactively detect pedestrians
users (VRUs) group, which includes pedestrian and cyclists, at risk. In Proceedings of the IEEE Vehicular Technology Conference,
pages 1–8, Sep 2012.
has the higher risks of traffic accidents. Recently, the use of [13] Doug Washburn, Usman Sindhu, Stephanie Balaouras, Rachel A. Dines,
smart computing technologies have been proposed to build Nick Hayes, and Lauren E. Nelson. Helping CIOs understand smart city
smart cities, as a strategy to mitigate the effects of the initiatives: Defining the smart city, its drivers, and the role of the CIO.
Forrester Research Inc., 2011.
problems of the growing urban areas.

312