Modeling, Design and Analysis of Intelligent

Traffic Control System Based on Integrated

Statistical Image Processing Techniques

Yasar Abbas UrRehmanl,2, Adam Khan I and Muhammad Tariq 2

I City University of Science and IT, Peshawar, Pakistan
2 FAST (NUCES) University Peshawar, Pakistan

Abstract-Traffic congestion is a growing problem in scene require description and discrimination from the
urban areas nowadays. In this paper, a system is developed background. This is normally called in literature as,
to control and monitor the congestion of traffic. The main defining Object of Interest (001). To efficiently control
motivation is to detect the presence and absence of vehicles the process, requires background modeling techniques
on the road using statistical approach integrated with
[1]. Other constraints are environmental variations such
conventional image processing techniques. For this purpose,
as illumination, fog, temperature, wind blows etc.
we have develop a "Probability Based Vehicle Detection
(PBVD)" algorithm based Vehicle Detection System (VDS)
Previous approaches process the whole two or three
integrated with post - processing subsystems to form a dimensional plane for object detection and isolation from
complete traffic control system. The system has the the background [2, 3]. Background subtraction for object
capability to obtain vehicle statistics during controlling detection was therefore the main constitute. However
traffic. Simulations are performed by developing complete background subtraction performs poorly in case of
prototype traffic architecture. Comparison is done using the outdoor environment where environmental variations
result acquired from prototype system and processing a real demand a robust background update algorithm [4].
time video of traffic scene. Simulation results show the
Ultimately, a vigorous background model need to adapt
effectiveness of the proposed scheme.
quickly to environmental variations and therefore it is a
I. INTRODUCTION separate research field in its own [5].
To tackle the aforementioned constraints, in this paper
Traffic control has been atopic of research from the
we propose an intelligent system for the control of traffic.
past few decades. Management of large traffic in
The system incorporates the integration of statistical and
metropolitan areas is not a trivial problem and requires a
digital image processing techniques. It is capable of
careful thought out planning and control. Beside,
controlling the traffic and at the same time gives us the
increasing road widths or making by - pass lines, the
road statistics. Moreover the system has the provision of
available technological assets can be optimized to
collecting data about the vehicles.
manage the intercity traffic. They allocate enough
The rest of the paper is structured as follows: Section II
resources toeffectively control the urban traffic.
discuss the related work. Section III presents problem
Additionally, their operation and maintenance cost are
description followed by vehicle detection in section IV.
low.
System design is elaborated in section V. Section VI
Traditionally, lamps are used to control the traffic at
shows experimental results and finally section VII
intersection. Every lane is allocated a fix timer in which
concludes the paper.
the vehicles are queued and when the timer expires
vehicles are allowed to pass by. There is limited II. RELATED WORK
intelligence involved in such systems. Magnetic loops
With the advancement in science and technology, It IS
were another method to detect the presence of vehicles on
now possible to monitor and control the large urban
the road.
traffic by optlmlzmg the available technological
Beside fix timers and magnetic loops there are other
resources. In the following discussion, some of the state
methods developed in the recent years for monitoring and
of the art work done for controlling traffic is presented.
efficient control of traffic. These involve the use of image
An intelligent system for the control of traffic light and
sensors and incorporating the techniques of digital image
density has been proposed [6]. The system uses IR
processing in the available hardware platforms. Image
(infrared) sensors for detection and microcontroller for
sensors can be used for diverse purposes such as:
controlling of light signals. However, this system poses
Monitoring vehicle flow on a given lane, calculating
certain limitations. First, infrared sensors have a very
speed, crash detection, Automatic Number Plate
short range. Secondly, most of the infrared light is
Recognition (ANPR) etc.
absorbed during the fog.
When controlling the intercity traffic using camera
A system for the estimation of traffic parameter using
controlled systems there are some constraints that must
mobile phone counts for measuring traffic density has
be taken into considerations: Detecting vehicles in the
been proposed [7]. In this method vehicles containing

978-1-4799-6369-0/15/$31.00 ©2015 IEEE

Proceedings of 2015 12th International Bhurban Conference on Applied Sciences & Technology (IBCAST) 169
Islamabad, Pakistan, 13th - 17th January, 2015
mobile phone entering and exiting the cells were counted.
Based on this counting a statistical measurement is made
about the number of vehicles. Bi is the background for ith frame Ii, Bo is the initial
An offline system for the estimation of traffic has been background and a control the rate of update.
proposed in [8]. In this method, a video film of road To detect vehicle on the road, a probabilistic approach
scene was captured and then analyzed frame by frame for beside the conventional approach of frame subtraction is
the existence and nonappearance of vehicles. An used:
indication of the heavy traffic was given once the number (2)
of vehicles exceeds the specified threshold.
where,
Traffic data collection has been done using visual
sensors. This type of data include: number of cars, their
Pi(k)
L is the probability that Lth intensity pixel having
intensity kL occur in the frame i, nL is the number of
sizes, orientation and appearance [9, lO]. A novel
occurrence of kL intensity pixel, N is the total pixels in
approach has been proposed for design of intelligent
an image, i.e. N m x n.=

traffic control system in [11]. Basically it's a hybrid

If there is some vehicle present on the road the number of
system in which mainly the traffic is controlled at the
pixels with Lth intensity will change from nL to n/ and
intersection, incorporating the techniques of image
processing and fuzzy logic. hence the probability from Pi(kL)to Pi+/ (kL) :
In [12], the author has developed a Mobile Intelligent
Traffic Control System (MITCS) for controlling intercity (3)
traffic in Taiwan.
The control of traffic flow based on Area of Interest
(4)
(AOI) using vehicle queue length and projection of two
dimensional coordinates on three dimensional planes has
been studied [13]. (5)

To simplify calculations, we make a use of histogram

III. PROBLEM DESCRIPTION and relate it with the above equations:
The problem description is two folds: First, given a
static camera mounted on the pole above the road, the (6)
task is to detect the absenceand appearance as well as the
number of vehicles if present on the road in a diverse (7)
environment with uncertain background. This is not a
trivial problem as it involves the information about the Thus:
lane region and background.
Secondly, an autonomous system is designed to detect, L - P·!(kL )1
Ip.!+1 I(k) -
-
I hi(kL)- h/(kL) I (8)
N
classify and display traffic on a given route. Furthermore,
the system is made to adaptively control the flow of
traffic in a given diverse and challenging environment. (9)
To tackle the first problem an adaptive background
modeling technique is exploited assuming that the road (10)
conditions vary linearly with time. Moreover, to detect
the presence and absence of vehicles a probabilistic
approach based on histograms of initial and current
The resultant value of T is compared with the
frames is used.
threshold "I" empirically calculated. If the value of T
In order to overcome the second problem, a complete
is greater than I it will indicate the presence of the
system architecture and algorithm is developed by using
vehicle in the scene and vice versa.
mathematical functions.
Since, the presence and absence of vehicle is similar to
IV. VEHICLE DETECTION tossing an unbiased coin therefore we define the
following function:
To detect the attendance and absence of vehicles on
the road, a description of vehicles feature and road profile
is required. The knowledge of background is the most
\1:.
! =
{I, T>t
(11)
0, else
crucial component in detection process. Moreover, lane
extraction will facilitate the detection of vehicles on the Having discussed about the vehicle detection the next
road. Since, in real time traffic applications we require section will illustrate the system design.
efficient result with less time; therefore background
modeling techniques are not appropriate. V. SYSTEM DESIGN
Let us consider that the camera is static and the
environment varies monotonically, we use the following
The system adopted is divided into four major
equation for background update:
components:

Proceedings of 2015 12th International Bhurban Conference on Applied Sciences & Technology (IBCAST) 170
Islamabad, Pakistan, 13th - 17th January, 2015
 A. Vehicle Detection System (VDS)
B. Vehicle Counting and Classification System Vehicle Detection ===::::> Vehicle Counting and classification
(VCCS) System System
C. Traffic Signals Control System (TSCS)
D. Data Display System (DDS) II H
Fig. 1 shows the overall operation of proposed Traffic Signal Control Data Display System
system. The vehicle detection system is use to monitor System

for the presence and absence of vehicles on the roads.

Fig. l System Design
The data collected by vehicle detection system is fed to
the vehicle counting and classification system. The
system analyzed the received data and accordingly makes
D. Data Display System
decision about the traffic signals. Also the data is sent to The Data Display System (DDS) display the total
the display device for displaying the road conditions, number of vehicles, the numberof pixels each vehicle
such as the number of vehicles, the number of pixels each contain, the number of vehicles fall into each category,
vehicle contain and classification of vehicles into small, and the priority given to road. In general any category for
medium and large size. In the following sections we will vehicle classification can be made however; we have
discuss the components of the system design. categorized the vehicles into three categories, small,
medium and large.
A. Vehicle Detection System
The Vehicle Detection System (VDS) is based on VI. EXPERIMENTAL RESULTS

Probability Based Vehicle Detection (PBVD) algorithm. The experiments were performed first in a standard
The algorithm starts with the initialization of control controlled environment with prototype traffic control
parameters followed by frame capturing operation. The system as shown in Fig.2. The cameras used in this work
capture frames are in true color format which is then are standard Logitech C110 cameras. The traffic was
converted to monochrome image. After, vehicle detection generated using remote control cars. At the beginning,
is triggered by calculating value T from the sum of both lanes lamp indications are red (to stop vehicles).
absolute histogram subtraction between the reference and When vehicle is detected by any of the camera it is sent
current frame. If T become greater than the empirically to VDS which uses PBVD algorithm to process and
predefined threshold value "t" then post processing is extract vehicles from the image. The parameter a. for
done to extract vehicles. This is done by subtraction of background update was set to 1 x 10-3, whereas the value
reference frame from the current frame followed by edge of threshold f for histogram comparison was set to 300.
" '

detection, binary thresh - holding, dilation, erosion and The processing is done using MATLAB® software for
labeling operations on the differential frame. both controlling the external cameras and serial
Afterward, area calculation is used to separate connected indication lamps.
vehicles from the redundant noise if present. Area Fig. 3 shows the DDS system for displaying road
calculation is more immune to noise as compared to other conditions and statistics.
techniques. Finally the detected vehicles are sent to Fig. 4 - 6 shows the complete process of extracting
Vehicle Counting and Classification System (VCCS) for vehicle from the scene using the prototype architecture.
counting and classification of vehicles. At the end of the As can be seen in Fig. 4 - 6, our prototype system gives
algorithm the reference frame is set accordingly and the accurate results even in the presence of noise. Since,
process is repeated. maximum area calculation does not involve any shape
distortion process; therefore the final results are very
B. Vehicle Counting and Classification System
efficient as can be seen in Fig. 6.
Data acquired from the VDS is labeled. After After testing the proposed scheme on prototype traffic
labeling, the numbers of pixels each labeled vehicle architecture, the system performance is then evaluated on
contain are counted and accordingly the vehicles are a real time video of highway traffic and compared both
categorized into small, medium and large vehicles. This the results. In processing this video we did not have any
data is then sent to the Data Display System (DDS). Also background information and thus it provide a challenging
in parallel, the numbers of labeled vehicles are counted task of object detection. This is shown in Fig. 7 - 11. Fig.
and sent to Traffic Signal Control System (TSCS). 7 shows the current frame. Fig. 8 shows the result after
C. Traffic Signal Control System frame subtraction. Fig. 9 shows the resultant image after
edge detection and binary dilation. Fig. 10 shows the
After receiving the vehicles statistics from VCCS, the
required image after binary erosion, labeling and area
Traffic Signal Control System (TSCS) compare the
calculation operations. Finally Fig. 11 shows the detected
number of vehicles on both sides of the roads and
objects. Fig. 12 depicted the spread resulted from
accordingly make a decision as to which side of the road
absolute histogram subtraction of the reference frame and
should be given priority. If both of the roads contain the
the current frame.
same number of vehicles then equal time will be given to
After comparing Fig. 4 - 6 with Fig. 8 - 10, it is noted
both lanes of the road.
that in real scenarios object shape is distorted during

Proceedings of 2015 12th International Bhurban Conference on Applied Sciences & Technology (IBCAST) 171
Islamabad, Pakistan, 13th - 17th January, 2015
processing of a real time video. This is mainly due to
edge detection and morphological operations (dilation,
erosion, hole - filling). Since the main work is focused on
detection and extraction of objects from the scene, we
noted that in both scenarios, the proposed system gives us
satisfactory results. Moreover, the area calculation
operation after labeling operation helped in extracting (a)
vehicles in the of noise.

(c) (d)
Fig. 5 a) Image after edge detection b) Binary conversion andbinary
dilation c) Binary Hole Filling d) Binary Erosion

(a)
Fig. 6 a) Area Calculation b) Final Result (Detected Objects)

(a) (b)

(c) (d)

Fig. 4 a) Image of empty road b) image of the road containing car c) Fig. 8 Frame subtraction
image enhancement after median filtering d) image after frame
subtraction

Proceedings of 2015 12th International Bhurban Conference on Applied Sciences & Technology (IBCAST) 172
Islamabad, Pakistan, 13th - 17th January, 2015
 REFERENCES

[I] G. Wang,D. Xiao and J. Gu, "Review on vehicle detection based

on video for traffic surveillance," In IEEE Int. Conf. on
Automation and Logistics", pp. 2961 - 2966, Qingdao, I - 3 Sept.
2008.
[2] S. Fazli, S. Mohammadi and M.Rahmani, "Neural Network based
Vehicle Classification for Intelligent Traffic Control," Int. J.
Software Engineering & Applications, Vol. 3, no. 3, pp. 17-22,
2012.
[3] M. Arora, and V. K. Banga, "Real Time Traffic Light Control
System Using Morphological Edge Detection and Fuzzy Logic,"
In2nd Int. Conf. on Electrical, Electronics and Civil Engineering
(ICEECE), pp. 28-29, Singapore April, 2012.
[4] J. Y. Hao, C. Li, Z. Kim and Z. Xiong, "Spatio-Temporal Traffic
Scene Modeling for Object Motion Detection," IEEE Trans. IntelI.
Transp. Sys., Vol. 14, no. I, pp. 295 - 302, 2013.
[5] S. Zhang, H. Yao and S. Liu, "Dynamic Background Modeling
and Subtraction using Spatio-Temporal Local Binary Patterns," In
IEEE Int. Conf. Image Processing (ICIP), pp. 1556 - 1559, San
Diego, CA, 12 - 15 Oct. 2008.
[6] A. MsPromilasinhmar, "Intelligent Traffic Light and Density
Control using lR Sensors and Microcontroller", International
Journal of Advanced Technology and Engineering Research
(IJATER), Vol. 2(2), pp. 30-35, March, 2012
[7] Vittorio Astarita, Roberto Bertini, Sergio d'Elia, Giuseppe Guido,
"Motorway traffic Parameter estimation from Mobile Phone
Fig. 10 Resultant image after binary hole filling, Counts", European Journal of Operational Research, Vol 175(3),
erosion and area calcuation pp. 1435-1446, December, 2006
[8] PejmanNiksaz, "Automatic Traffic Estimation using Image
Processing", International Journal of Signal Processing, Image
Processing and Pattern Recognition, Vol 5(4), pp. 167-174,
December, 2012
[9] C. Mallikarjuna, A. Phanindra, K. RamachandraRao, "Journal of
Transportation Engineering", Vol. 135(4), pp. 174-182, April
2009
[10] P. Reinartz, M. Lachaise, E. Schmeer, T. Krauss, H. Runge,
"Traffic Monitoring wit Serial Images from Airborne Cameras",
ISPRS Journal of Photogrammetry and Remote Sensing , Volume
61(3-4), pp 149-158, December 2006
[11] O. UchennaChinyere, O. Onaolapo Francisca, O. Emmanuel
Amano, "Design and Simulation of an Intelligent Traffic Control
system", Int. 1. of Advances in Engineering & Technology, Vol.
1(5) pp.47-57, Nov 2011
Fig. II Final result
[12] L. Tay Lin, H. Jen Huang, 1. Min Lin, F. Frang Young, "A New
Intelligent Traffic Control System for Taiwan," In 9th Int. Conf.
VII. CONCLUSION
Intelligent Transportation Sys. Telecommunications (ITST), pp.
In this paper, an intelligent traffic control system is 138 - 142, Lille, 20 - 22 Oct. 2009.
proposed for monitoring and control of intercity traffic. [13] C. Yingfeng, Z. Weigong, W. Hai, "Measurement of Vehicle
It is then tested using prototype traffic architecture and Queue Length Based on Video Processing In Intelligent Traffic
by processing a real time video of a high way. The Signal Control System," In In!. Conf. Measuring Tech.

final results are satisfactory and show that the system Mechatronics Automation (ICMTMA), pp. 615 - 618, Changsha
City, 13 - 14 Mar. 2010
can cope with a noisy environment. Furthennore, we
found that morphological operation greatly distort the
shape of object and so there is some need of
restoration techniques to retain the object shape after
final detection.

Proceedings of 2015 12th International Bhurban Conference on Applied Sciences & Technology (IBCAST) 173
Islamabad, Pakistan, 13th - 17th January, 2015
 Spred after absolute histogra1TI subtraction
30

25

20
<U
�
� "
'6
E
15
='

�
-<
.=

10

.,
• .,

5 Itl
,"'I
� a .. 11
"'I
II
r
WlIU
u:
�. t...llit.� �ft
50
rw VI r ��
'f )0
.,

Pixel ll"1tensities
150
- M� � i �.4 1',--
20 0
:41
EMt
:'Y'-.Ir'
2�(Y

Fig, 12 Spread after absolute histogram subtraction

Proceedings of 2015 12th International Bhurban Conference on Applied Sciences & Technology (IBCAST) 174
Islamabad, Pakistan, 13th - 17th January, 2015