2019 IEEE 4th International Conference on Computer and Communication Systems

Real Time Adaptive Traffic Control System

A ybrid Approach

Prayushi Faldu Nishant Doshi

Information and Communication Technology, Computer Science and Engineering,
Pandit Deendayal Petroleum University, Pandit Deendayal Petroleum University,
Gandhinagar,India, Gandhinagar,India
e-mail: [email protected] e-mail: [email protected]

Reema Patel
Computer Science and Engineering,
Pandit Deendayal Petroleum University,
Gandhinagar,India
e-mail: [email protected]

Abstract—In present world, the problem of traffic congestion The main cause of traffic congestion is found to be lack
has become a major concern. It’s not just restricted to of a dynamic traffic control system with proper prioritization.
megacities or metropolitans but even for small cities. Hence we The present traffic control system have fixed green path
require a smart or intelligent traffic control system. Our clearance time which is completely independent of the traffic
present traffic control system is not adaptive but is based on density. Even if the number of vehicles on a particular lane is
time and independent of the traffic density. This static nature less the green signal time is static and vehicles of other lanes
makes it inefficient, chaotic and noisy. To tackle this issue of have to wait unnecessarily. In addition to this there no
managing traffic roads a real time traffic information prioritization for emergency vehicles which can be proved
collection and monitoring system is proposed. The proposed
life threatening. Hence this has become very important to
model controls the clearance time of each lane in a sequential
manner and is function of real time traffic density. The
solve this problem.
approach is rather hybrid – a blend of sensors networks and The problem can be solved by using present emergent
camera technology. United results of image processing and technologies like IOT and image processing. A large amount
IOT ensures more accuracy and efficiency rather than native of research is going on in these fields. In our proposed
approach. Besides traffic congestion minimization, system we have made use of these technologies in order to
prioritization for emergency vehicle is also achieved through develop a smart and intelligent traffic control system. Both
RFID. Other technology like stolen vehicle tracking/detection of the methods individually have their own drawbacks
is also deployed. Simple GUI for use add an advantage to the individually. The proposed model hence uses a hybrid
system making it easy for control room to handle. The model approach to provide a better and efficient solution to the
could be used for analysis purpose and hence predict traffic problem hence overcoming the problems faced. Formula for
jams and traffic status at different times. Hence proposed calculating green clearance time based on individual is also
model is cost-effective and time efficient smart traffic control discussed further. In addition to this the paper also proposes
system. models for emergency vehicle clearance and stolen vehicle
detection. The model can be further extended to predict the
Keywords-component; image processing, congestion control, traffic jam.
sensor networks, raspberry pi, smart traffic control system,
hybrid approach, emergency prioritization, stolen vehicle II. LITERATURE SURVEY
detection
Various technologies have been proposed so far for smart
traffic control systems. These include a wide range of
I. INTRODUCTION technologies like Traffic warning system, Zigbee, ultrasonic
In growing metropolitans across the world, traffic sensors, RFID, Camera surveillance, load cells, queue
congestion have become an inescapable situation. This detectors, GPS, wireless Sensor Networks (WSN) [3][4][5].
increasing amount of traffic is a result of increasing number To increase the number of vehicles passing in the green
of vehicles and lack of a proper system to handle it. Increase path an algorithm was proposed. Gist of the algorithm was
in traffic congestion even leads to wastage of fuel, time and priority number. Each and every vehicle was assigned a
hence increase the overall transportation cost[1].This priority number. Different classes of vehicles can be
unavoidable problem is faced by almost every country which emergency vehicles like ambulance, fire brigades are given
have other indirect problems associated with it[2]. the highest priority, then comes the vehicles of VIP’s or

978-1-7281-1322-7/19/$31.00 ©2019 IEEE 697

government officials, and then the ordinary vehicles used by task of density detection. Pi-camera is used to capture images
us. Prioritization was done based on vehicle density [6]. which is then processed using openCV to find traffic density.
Another method was proposed which used inductive In addition to it RFID is used to detect entry of any
looping technique. To measure the traffic density on emergency vehicle like fire brigade or ambulance. The
different lanes a microcontroller was designed which proposed model also reports the detection of stolen
collected the information from inductive loops and gives the automobiles, which can be again done by RFID.
output accordingly [7]. The complete prototype can be divided into three
A smart highway system was also proposed which modules – congestion control, emergency vehicle detection
involved measuring vehicle speed, traffic density, vehicle and reporting of stolen automobiles. One of the major
routes and pattern of traffic at different times in different drawback of existing traffic control system is that it is static.
routes. IR led and photo detectors were used to detect and The proposed module gives green path time which is a
evaluate the traffic flow [8]. function of traffic density. Traffic density is calculated by
A smart traffic management system was developed which using a hybrid approach i.e. IR sensors and Image processing.
used load cells for traffic density. But these load cells are A pair of IR sensor is used for this purpose. One is IR
needed to be deployed on the road i.e. beneath the vehicles. transmitter and other is IR receiver. IR transmitter emits
They are programmed such that they calculate the traffic Infrared rays. IR receiver is a photodiode which is
density based on pressure and position of vehicles [9]. responsible for receiving the reflected IR rays and remains in
inactive stage. The mechanism is as follow: IR transmitter
III. PROPOSED MODEL generates a square wave of 38 kHz continuously which gets
Various technologies have been proposed so far for real reflected if presence of any object is detected, these reflected
time adaptive traffic control system. Technologies like rays are then received by the IR receiver and the system is
inductive looping and load cells requires in ground activated. The output is HIGH i.e. 3.3V if an object is
installation which increases deployment and maintenance detected or else it is LOW i.e. 0V. IR sensors will be placed
cost. Individual systems have drawbacks which are needed to at some regular intervals in each lane of the traffic signal. Let
be solved to improve the efficiency of the system. In d meter be the distance at which first IR sensor is placed.
proposed system we have provided solution to adaptive Now, we will place rest three sensors at some interval let’s
congestion control mechanism. In addition to it clearance say p meters. So second sensor is placed at a distance of (d+
prioritization to emergency vehicles and stolen vehicle p) meters, third at (d+2*p) meters and the last at (d+3*p)
detection mechanism is also proposed. The model is meters. Since output of sensor is in binary form we can have
designed so as to provide a low cost and power efficient following table TABLE.1 for deciding about the mode of
solution to the system. traffic.
The main advantage of this model is increased efficiency This is the formula used to calculate the index of the
due to hybrid approach. Using both technologies individually counter.
have their own drawbacks. Fluctuations in light intensities or
weather conditions may result in inaccurate calculation of Index [ i ] = Output [ i + 1 ] AND Output [ i ]
traffic density by image processing technique if used alone.
Deployment of only IR sensors may be inaccurate in the Here AND is the Boolean operator which results into 0 if
cases when there is a significant amount of gap between the any of the input value is zero and gives output as 1 only
vehicles in a particular lane. Hybrid approach helps to when both of the input values are 1.
overcome their drawbacks by considering the weighted
average of both results hence resulting into better results with TABLE I. FOR IR SENSOR
more accuracy. For other existing technologies and SENSORS Output Index
methodologies failure of any one of the method will lead to
SENSOR 1 ( S1 ) 1 or 0 S1 AND S2
failure of system if used individually unlike hybrid approach
where one system will always be active in case of any SENSOR 2 ( S2 ) 1 or 0 S2 AND S3
system failure.
SENSOR 3 ( S3 ) 1 or 0 S3 AND S4
Image processing is done on the image captured by the pi
camera which is placed at the intersection of four lanes, SENSOR 4 ( S4 ) 1 or 0 S4 AND 1
hence camera rotates 90 degree after every green signal.
Image captured will be then processed to calculate traffic We categorize the traffic into four different modes on the
density. At the same time the IR sensors which are deployed basis of density of the traffic calculated: Extremely High
on different lanes of road also collect traffic density traffic, High traffic, Medium traffic and Low traffic. Traffic
information. These values are then combined to find the density will be calculated on basis of first zero found at any
cumulative density and hence clearance time or green path index. If index is zero at S1 then traffic density will be low,
time. if first zero at S2 then medium and so on. This was about
Proposed system is shown in Figure 2 Here Raspberry pi sensors.
is our intelligent IOT gateway which performs the task of At the same time an image will be captured by the pi
processing. We have used IR sensors for traffic density camera installed at the junction/intersection. Processing of
detection. A pair of transmitter and receiver performs this

698
the image in divided into different steps which is carried out hybrid operators open which is erosion followed by
using openCV [10][11][12][13] Figure 3. Shows output at dilation and close which is dilation followed by
different stages. Different stages are as follow: erosion. Open removes all white dots except
1. Preprocessing: Captured image first undergoes vehicle and close removes all black dots except
some pre-processing steps. Basically it consist of
background.
two sub stages i.e. Gray scale conversion and
After all of these stages blob analysis is carried out
image enhancement. Image enhancement is a
preprocessing method to enhance the image which detects the traffic density and sends the
features like brightness and contrast. Method used result to the raspberry pi.
for this is power law transformation. It relates input Now we have results from both sensor and camera
image with output image by a multiplication factor which can be used to find the cumulative density.
gamma. Proper gamma selection is an essential We will use following mapping table to calculate
because fineness of output depends on it. Gamma individual clearance time.
for our system is “1.5”. Another sub stage is gray
scale conversion. Here color image is converted to
TABLE II. MAPPING TRAFFIC DENSITY AND WEIGHTS
gray image by transforming the R, G, B values of
each pixel. The output of this will be a gray Traffic Mode Clearance Time Weight
enhanced image. Low 15 1
2. Image subtraction: Our main aim to detect the
Medium 30 2
vehicles from the image captured and hence find
the traffic density. Vehicles are the moving or High 45 3
dynamic objects. And the road behind is static i.e. Extremely High 60 4
remains the same even with the changing traffic.
This idea is used here to identify the dynamic
objects. Approach is to compare image with a Let CT1 be clearance time calculated by IR sensor and
CT2 be clearance time calculated by camera. W1 and W2 be
reference image. So, now we have two components
respective weights. Now the final green path /clearance time
– foreground and background. We will separate is calculated as weighted average of the individual values.
two layers by subtracting background image from Below formula gives us the net green path time which is
captured image to get the foreground image. The fed as input to the the green signal.
output will be those pixels which are different from
the background image (reference image). Hence
the input will be captured image and background
image will be the empty road. And output will be
frames which contains the vehicles (dynamic
objects).
3. Gaussian filtering: It is one of the small but useful
stage. This step is essential before image
binarization as it improves output efficiency of it.
The input image is filtered by convolution of each
point/pixel in image with the Gaussian kernel. And
the output is summation of all results of
convolution.
4. Image Binarization: After Gaussian filtering now
we convert the image into black and white i.e.
image in binary colors. It is achieved through
thresholding. We have used Otsu’s Binarization
method for this. It gives the best result because best
suitable threshold value not known here. The input
is gray image and output will be image where all
vehicles will be white and background as black.
5. Morphological transformations: It is similar to
image filtering but this time the operation is on the
shape of image. A total of two inputs are there –
binary image and structuring element. It uses two
Figure 1. Traffic control system architecture

699
 Figure 1. shows architecture for traffic control system. vehicles have RF transmitters. Now the communication will
In addition to congestion control, the proposed model take place between this two. As soon as RF receiver detects
also comprises of priority clearance for emergency vehicles. the presence of emergency vehicle it immediately turns green
It is detected using RF receivers deployed somewhere near signal on. The signal remains green till receiver receives
sensor 3 and at the junction too. We know that emergency signal.

Figure 2. Prototype for proposed model

Figure 3. Various stages of image processing

700
 The third module is stolen vehicle detection. This can be [2] MihaelaCondurat, Alina Mihaela Nicuţă, RaduAndrei,
again implemented with use of unique RFID tags. RFIDs of “Environmental Impact of Road Transport Traffic. A Case Study for
County of Iaşi Road Network”, 10th International Conference
the stolen vehicle can be stored in the database which can be Interdisciplinarity in Engineering, Procedia Engineering, Volume
further compared by RFID reader. If a RFID reader detects 181, 2017, Pages 123-130.K. Elissa, “Title of paper if known,”
some stolen vehicle then immediately signal could be turned unpublished.
red and a notification is send to the control room containing [3] H. O. Al-Sakran, "Intelligent traffic information system based on the
details of the stolen vehicle and RFID number. integration of Internet of Things and Agent technology," International
Journal of Advanced Computer Science and Applications (IJACSA),
IV. CONCLUSION AND FUTURE ENHANCEMENTS vol. 6, no. 2, pp. 37-43, 2016
[4] Bilal Ghazal, Khaled EIKhatib, Khaled Chahine, Mohamad Kherfan,
The proposed model provides solution to the growing "Smart traffic Light Control System" , IEEE 2016.
traffic congestion problem and can effectively replace the [5] Mohammad S. Jassas, Abdullah A. Qasem, Qusay H. Mahmoud, “A
existing traditional methodologies or traffic control system. Smart System Connecting e-Health Sensors and the Cloud”,
Since the green pathh time is a function of the traffic density Proceeding of the IEEE 28th Canadian Conference on Electrical and
it could be varied with the varying traffic densities and Computer Engineering Halifax, Canada, May 3-6, 2015.
conditions. Clerance time calculation is adaptive and [6] Abishek C, Mukul Kumar and Kumar Padmanabh "City Traffic
intelligent, in addition to this prioritization for emergency Congestion Control in Indian Scenario using Wireless Sensors
Network".R. Nicole, “Title of paper with only first word capitalized,”
vehicles and detetction or tracking of the stolen vehicles J. Name Stand. Abbrev., in press.
makes it more smart. The hybrid approach used in the [7] Nayak, R. R., Sahana, S. K., Bagalkot, A. S., Soumya, M., Roopa, J.,
proposed model creates an edge over other methods. The Govinda, R. M., & Ramavenkateswaran, N. (2013). Smart traffic
data is collected at both local and central level hence makes congestion control using wireless Communication. International
it better for situation when one of them fails. The cumulative Journal of Advanced Research in Computer and Communication
Engineering, 2(9).
density calculated gives precise clearance time since both the
individual methods have some drawbacks in certain [8] Integrating IR Sensors with a Microcontroller. Automated Regulation
of Traffic Flow”, Robert H. Winston.Electronic Publication: Digital
conditions, so the hybrid model proposed tries to reduce Object Identifiers (DOIs):
errors in one method and hence provide more accurate [9] Chandana K K, Dr. S. Meenakshi Sundaram, Cyana D’sa, Meghana N
results.All the data collected and status of the traffic is Swamy, Navya K, “A Smart Traffic Management System for
reported to the control room and could be accessed by simple Congestion Control and Warnings Using Internet of Things (IoT)” ,
GUI. These could be further extended to show traffic status Saudi Journal of Engineering and Technology, Vol-2, Iss-5(May,
to the users. The data colleceted of traffic density at different 2017):192-196 .
times of day can be used for analysis purpose and prediction [10] Ali Tourani, Asadollah Shahbahrami, Vehicle Counting Method
Based on Digital Image Processing Algorithms, 2nd International
of the traffic at different times of a day. The analysed data Conference on Pattern Recognition and Image Analysis (IPRIA 2015)
then could be used to predict traffic jams at different March 11-12, 2015
locations of the city. Hence the proposed model is a [11] Priyanka Khanke, Prof. P. S. Kulkarni , "A Technique on Road Tranc
complete package for smart traffic sontrol system which Analysis using Image Processing", Vol. 3 Issue 2, February 2014
could be extended for making complete trasportation system [12] de-la-Rocha, E.; Palacios, R.; "Image-processing algorithms for
intelligent. detecting and counting vehicles waiting at a traffic light", Journal of
Electronic Imaging, vol.19, no.4, pp.043025-1-043025-8. October,
REFERENCES 2010
[1] Sabya sanchi kanojia, “Real –time Traffic light control and [13] de-la-Rocha, E.; Palacios, R.; "Image-processing algorithms for
Congestion avoidance system”, International Journal of Engineering detecting and counting vehicles waiting at a traffic light", Journal of
Research and Applications (IJERA), pp.925-929, Vol. 2, Issue 2,Mar- Electronic Imaging, vol.19, no.4, pp.043025-1-043025-8. October,
2010.
Apr 2012J. Clerk Maxwell, A Treatise on Electricity and Magnetism,
3rd ed., vol. 2. Oxford: Clarendon, 1892, pp.68–73.

701