Real Time Adaptive Traffic Control System for Smart Cities

CHAPTER 1

INTRODUCTION
In country like India, billions of people start and end each working day stuck in traffic
or commuting on congested trains and buses. Traffic has been a routine problem for
everyone. There are several reasons for traffic congestion such as massive expansion of
vehicles due to growth in population, poor infrastructure, poorly controlled traffic but the
main reason is slowed synchronization of traffic signal. Traffic congestion is leading to
reduced productivity, air pollution, fuel wastage and accidents . Therefore traffic congestion
must be reduced.

It is vital to the quality of life to enhance the everyday commute. By 2025, cities that
implement smart mobility systems on average, reduce commuting cycles by 15-20 percent,
with some individuals experiencing even greater reductions. Depending on each city's
density, current transit facilities, and commuting habits, the capacity associated with each
application is highly variable. Slowed synchronization of traffic signals leads to traffic
congestion and delays. The pre-programmed, regular signal timing patterns are employed in
traditional signal systems.

To overcome the problems of traditional traffic control systems, there is a shift in adaption to
an Adaptive traffic control system. The Adaptive Traffic Control System (ATCS) is a traffic
management technique that modifies or adapts the timing of traffic signals based on the real
demand for traffic and achieved using a control system that includes both hardware and
software, where hardware is the sensor used for real-time traffic density estimation and
software is designed using captured data analysis of the city's current traffic flow. This paper
depicts a model of camera-based traffic monitoring and processing system which reduces the
cycle time and possesses special provisions for emergency vehicles.

Congestion of traffic has been a matter of routine in any existing metropolitan region with an
overwhelming population but minimal infrastructure. Traffic Congestion is a critical problem
with dire causes and consequences on the road. Radical population growth and low public
quality transportation have caused vehicles to expand massively. Poorly controlled traffic,
apart from infrastructure, creates congestion that could survive for hours. Only to a certain
degree can a pre-defined timing scheme for traffic control ease the problem. It must surpass

Dept of ISE 2022-23 1|Page

Real Time Adaptive Traffic Control System for Smart Cities
its value as a fixed delay unit irrespective of the traffic volume on the pre-defined timer
signal that leads to more traffic build-up on other lanes of the intersections.

Many systems for traffic control handle signals on a fixed- time basis, where a sequence of
protocols for signal timing is defined by the day of the week and time in the day. The time
correlation is pre-calculated between the signals based on traffic circumstances previously
analyzed at particular junctions. In these fixed time models, it is not possible to expect
systems to cope with traffic conditions that vary periodically from those that existed at the
time when the junction was investigated.

In addition, as the patterns of traffic change dynamically, fixed-time techniques for traffic
signals often become redundant. This includes resurveying the region and fresh new signal
timing plans must be determined. Perspective has shown that this technique is extensive and
that resources that are not always readily available are necessary. The problems of most fixed
time systems make it clear that a more responsive approach to changing traffic conditions is
needed. The adaptive traffic control system is one such efficient solution in terms of traffic
management. This is a great change compared to fixed time structures because it means better
decision-making capabilities.

Dept of ISE 2022-23 2|Page

Real Time Adaptive Traffic Control System for Smart Cities

CHAPTER 2

EXISTING SYSTEM
The Existing System for traffic control is either manually controlled by traffic police
or automatically controlled by fixed timings technique. Here, the traffic signal timings is pre-
calculated between the signals based on traffic circumstances previously analyzed at
particular junctions. These timings are loaded into the timer of the traffic light which
automatically changes the traffic signals according to the fixed timer. The fixed timing
technique does not work as the traffic condition changes both periodically and dynamically.

Poorly controlled traffic, apart from infrastructure, creates congestion that could survive for
hours. Only to a certain degree can a pre-defined timing scheme for traffic control ease the
problem. It must surpass its value as a fixed delay unit irrespective of the traffic volume on
the pre-defined timer signal that leads to more traffic build-up on other lanes of the
intersections.

Many systems for traffic control handle signals on a fixed- time basis, where a sequence of
protocols for signal timing is defined by the day of the week and time in the day. The time
correlation is pre-calculated between the signals based on traffic circumstances previously
analyzed at particular junctions. In these fixed time models, it is not possible to expect
systems to cope with traffic conditions that vary periodically from those that existed at the
time when the junction was investigated.

In addition, as the patterns of traffic change dynamically, fixed-time techniques for traffic
signals often become redundant. This includes resurveying the region and fresh new signal
timing plans must be determined. Perspective has shown that this technique is extensive and
that resources that are not always readily available are necessary. The problems of most fixed
time systems make it clear that a more responsive approach to changing traffic conditions is
needed . The adaptive traffic control system is one such efficient solution in terms of traffic
management.

Dept of ISE 2022-23 3|Page

Real Time Adaptive Traffic Control System for Smart Cities

CHAPTER 3

PROPOSED SYSTEM
The fundamental concept behind having an adaptive traffic control system is to
construct a traffic management system in such a way that it functions in response to real-time
traffic patterns. We also come across existing traffic light system situations where, if there is
no vehicle on a specific lane and heavy traffic on another lane, for a specified time, the traffic
signal on the latter lane is green while the other remains red at the very same time. The
solution to these problems is to provide an adaptive system that works in accordance with the
number of vehicles lanes. This traffic system counts the number of vehicles in the lane,
compares the data from other lanes, and, based on the number of vehicles, determines
whether to offer priority to reducing the average waiting time or reducing
congestion/pollution in the lane.

The use of a camera-based ATCS system not only offers the solution for traffic congestion,
but it also provides additional features for intelligent transport systems like violation
detections, prioritizing emergency vehicles, road surveillances, and better civic management.
The adaptive traffic control system has become part of intra-urban transportation systems in
several cities in developing countries. It is part of the new generation of signaled intersection
control.

3.1 Intelligent Transport System

Intelligent Transport System (ITS) Architecture offers a framework for any form of
technology-related initiative in the transport sector. The system architecture is a broad
description of the facilities, ITS modules, interconnections, and information flow mapping for
the different systems and subsystems that cover the ITS project. With a properly designed
infrastructure, owners and stakeholders may define both the services provided by end-users,
the data for these services, and the interconnection between the various sub-systems.

The preparation and implementation of ITS are in their infancy in India. There are several
noteworthy achievements in ITS applications in India mainly in transport technology, traffic
control, fare collection, and toll collection systems. However, there is currently no National
ITS Plan or ITS Architecture that has been established. The lack of a national ITS
Dept of ISE 2022-23 4|Page
Real Time Adaptive Traffic Control System for Smart Cities
architecture does not mean that this move is not necessary. In fact, the World Bank
recommends that for those countries and regions that have not yet developed a national or
regional ITS Architecture, the ITS Architecture Project Level is a suggested initial step
towards starting the process of developing a national ITS Architecture.

Figure: 3.1 Layers of Intelligent Transport System

The ITS architecture as a whole involves many user services like Traveler information
services, Traffic management services, Public transport services, Electronic payment services
for Toll, Commercial vehicle operations, Emergency management services, Vehicle safety,
and control system and Information warehousing services. In the metro cities of India, the
existing traffic control system is inefficient because of randomness in the pattern of traffic

Dept of ISE 2022-23 5|Page

Real Time Adaptive Traffic Control System for Smart Cities
density during the day. Because of this, even though the traffic level is much smaller,
vehicles must wait for a long time. Hence, the traffic congestion problem can be reduced if
the traffic signal timer configured to be manipulated with a constant varying traffic density.

3.2 Adaptive Traffic Control System Framework

The ATCS algorithm continuously changes traffic signal timings based on the demand for
traffic at the intersections and projected arrivals from neighbouring intersections. It greatly
enhances travel time by pushing vehicles steadily through green lights and reduces
congestion by creating smoother flow.

Figure: 3.2 Overview of ATCS framework

Dept of ISE 2022-23 6|Page

Real Time Adaptive Traffic Control System for Smart Cities

Using image processing, the proposed system adapts the traffic signal timer according to the
random traffic density. This model uses high-resolution cameras to detect shifting traffic
patterns around the traffic signal and to manipulate the signal timer accordingly by sending
signals to the timer control system.

Traffic congestion depends directly on the regulation of the traffic flow and thus on the timer
of the traffic signal. As a result of this phenomenon, vehicles have to face an irregular pause
during travel in metropolitan areas. At currently, traffic control systems in India lack
intelligence and function as an open-loop system with no feedback or network sensing. The
goal of this work is to enhance the traffic control system by adding a sensing network that
provides input to the existing network so that it can respond to evolving patterns of traffic
density and provide the controller with the required signals in real-time. Our key aim is to
optimize vehicle transit delays in irregular hours of the day.

The approach suggested in this work consists of segmenting the problem into two sub-
problems, first of which is to calculate traffic volume using classifiers under the Viola Jones
Object Detection System. Using OpenCV libraries in the C/C++ environment, real-time
traffic volume estimation was obtained. In contrast to other libraries, OpenCV was selected
for time- efficient and reliable performance in real-time projects. The second sub-issue
included the development of a control algorithm to overcome traffic congestion.

Dept of ISE 2022-23 7|Page

Real Time Adaptive Traffic Control System for Smart Cities

Figure: 3.3 Sequence of steps for ATCS

CHAPTER 4

IMPLEMENTATION
To optimally overcome traffic congestion, decisions to trigger traffic signals need to
be based on real-time results. The following steps were followed as part of adaptive traffic
control.

A. Video feed acquisition

The control algorithm is intended to control 4-way and 2- way approach intersection. This
includes taking video data from various cameras and manipulating the input by suitable filters
and algorithms. The video feed is obtained via the automated operations supported by
OpenCV using the Direct Display (Dshow) techniques. The use of Dshow helps us to process
multiple feeds at comparatively higher speeds compared to other techniques. The camera
feeds are processed in the Integrated Control Center with the aid of big data as a backend
technology.

Dept of ISE 2022-23 8|Page

Real Time Adaptive Traffic Control System for Smart Cities

Figure: 4.1 Flowchart of resolution of traffic congestion by OpenCV

B. Pre-processing of video

Video processing is achieved by the repeated trials of the techniques of image processing to
video frames. Video feeds from various cameras are collected and analyzed frame by frame
in real-time, and the loop is iterated before the user wishes to interrupt it. Video pre-
processing methods are used to enhance efficiency and precision in the calculation of traffic
flow. It is also a mixture of multiple operations. We used two methods to achieve two
separate goals.

Firstly, the video stream is transformed from the RGB color mode to the Gray scale. Three
channel colors are translated to matching gray intensities. This stage decreases the volume of
data to be processed and thus improves the speed of process. In the second stage of pre-
processing, all video frames are equalized Histogram.

Histogram equalization results in an image with significantly higher contrast and does so by
extending the distribution of the size of the image. The histogram represents the pixel

Dept of ISE 2022-23 9|Page

Real Time Adaptive Traffic Control System for Smart Cities
intensities of the image with the corresponding values in the chart. A better outcome is
reached when the picture is first transformed to a gray scale and the histogram is equalized as
it increases the probability of an object being recognized.

Figure: 4.2 Image processing by OpenCV

C. Recognition of object

The classifier for the identification of objects, which in this case are vehicles, has been
created using the Open CV development kit. The cascaded classifier, consisting of several
stages of other simple classifiers, uses Haar-like features to detect objects. The classifier was
trained with 500positive samples and 1000 negative samples of vehicle images. The classifier
has been checked on the research samples and the outcome are seen in the below table.

Dept of ISE 2022-23 10 | P a g e

Real Time Adaptive Traffic Control System for Smart Cities

Table 4.1: Performance of Classifier for Object Detection

The real time output of the system is presented in the GUI of the ICCC platform for real time
monitoring of the traffic and to extract beneficial information for decision making by
authorities.

Figure: 4.3 Recognition of vehicles in OpenCV

D. Extraction of data

Data Extraction refers to the acquisition of data/information that is most suited to our
specifications from pre-processed video streams. Many variables can be derived using
various techniques. In case of object detection, the related variables may be Motion, Color,
Size, Object Centre, Direction, Total Count, etc. For our analysis, we only needed the total
number of items observed, which will apply to the number of vehicles in the queue at the
traffic intersection approach. The cumulative number of vehicles recognized shall be taken

Dept of ISE 2022-23 11 | P a g e

Real Time Adaptive Traffic Control System for Smart Cities
into account. The count for each approach of the intersection shall be fed to the algorithm for
the determination of the appropriate action.

Figure: 4.4 Extraction of data (Vehicle count) in OpenCV

E. Traffic control algorithm

The control algorithm is responsible for controlling traffic at the intersection when is supplied
with data. The algorithm was coded explicitly in the C/C++ interface to make it more
effective and stable. The algorithm complies with the following simple principles for the
management of traffic signals used in conjunction to minimize traffic congestion at the
intersection.

The control algorithm is based on what area the vehicles are occupied frame. If traffic is
greater than 80% of the total area, Read Traffic using camera then maximum time tmax is
allotted to the lane. If the traffic of the lane is less than 25% then minimum time tmin is
allotted to the lane. Time limits shall be maintained to prevent the arrest of the system in one
situation. The final basic rule aims to guarantee that all methods are granted an equal amount
of opportunities for optimal use. The emergency vehicles are given a priority access, so that
when the vehicles are detected, the system is overridden by special request from the
subroutine and the lane which contains the emergency vehicle such as ambulance, police cars
and fire trucks are given priority. Then, the flow switches back to the main routine. The

Dept of ISE 2022-23 12 | P a g e

Real Time Adaptive Traffic Control System for Smart Cities
dynamic timing control along with traditional way of queue-based priority method eliminates
the shortcomings of the previous methods and enhances the efficiency by reducing the cycle
time.

CHAPTER 5

RESULT

Dept of ISE 2022-23 13 | P a g e

Real Time Adaptive Traffic Control System for Smart Cities
Traffic simulation is achieved by using a framed algorithm in a 4-way path with 100
vehicles which spawn in random lanes. The size of the road is taken as 7 meters per lane and
the size of the vehicles are taken as 3.5 meters x 1.5 meters.

The speed of the vehicles crossing the junction is taken as 20 kmph. The Python 2.7 programs
were backed by the OpenCV image processing software and its associated libraries. The
picture is read from the video and translated to the color saturation and value (HSV) form.
The background is subtracted from the picture to get the image of the object of interest. The
picture is transformed to binary and the subject area is measured. Appropriate traffic has been
implemented by several cycles at each terminal of the junction.

Figure: 5.1 Simulation of ATCS to calculate cycle time

The table shown reflects the four-way intersection timing cycle. The first two sets are the
traditional methods that is programmed in fixed time traffic control systems. The last two
cycles have been programmed for the ATCS system.

Dept of ISE 2022-23 14 | P a g e

Real Time Adaptive Traffic Control System for Smart Cities

Table 5.1: Timing Diagram of the Simulation of the 100 Vehicles

In set A – conventional approach, the 100 vehicles are spawned in random lane and the traffic
lights operate in fixed timing mode. So, the overall time to resolve 100 vehicles in for set A
simulation is around 240 seconds.

In Set B, the overall time is 160 seconds in which the control algorithm employs pre-
surveyed data of each lane. In set A & set B simulation when ATCS is programmed the total
time of resolving the congestion for 100 vehicles is 105 seconds, which is very much less
than the traditional approaches. It is noted that the suggested algorithm resolves congestion
within a shorter span of time.

CONCLUSION

Dept of ISE 2022-23 15 | P a g e

Real Time Adaptive Traffic Control System for Smart Cities
The key benefit of the ATCS framework is that it is stable and easy to install. Using image
processing technique, the calculation of the length of the queue across junctions can prove to
be more accurate than conventional inferential sensing devices and infrared sensors. The
ATCS system also help to reduce the level of noise and air pollution, resulting in low fuel
consumption as well. This system shows that it can stabilize traffic in 4-way intersection
squares in far less cycles than in traditional fixed time timers with less energy and
maintenance. There is an immense number of developments in the modern form of an
adaptive traffic control system which might include AI based image processing. The system
can be significantly enhanced with the introduction of techniques such as solar operated
devices and sensors, LED display boards at junctions, synchronization of all traffic signals in
the city and high- resolution cameras with enhanced night vision capabilities.

REFERENCES

Dept of ISE 2022-23 16 | P a g e

Real Time Adaptive Traffic Control System for Smart Cities

[1] M. Sullman and L. Dorn, Advances in traffic psychology: Human factors in Road and
rail Transport. Ashgate, 2012.
[2] Studer L, Ketabdari M, Marchionni G (2015) Analysis of Adaptive Traffic Control
Systems Design of a Decision Support System for Better
Choices. J Civil Environ Eng 5: 195
[3] http://www.mohua.gov.in/upload/uploadfiles/files/OPERATIONS_TM ICC.pdf
[4] N. Rida, M. Ouadoud, A. Hasbi and S. Chebli, "Adaptive Traffic Light Control
System Using Wireless Sensors Networks," 2018 IEEE 5th International Congress on
Information Science and Technology (CiSt), Marrakech, 2018, pp. 552-556, doi:
10.1109/CIST.2018.8596620.
[5] https://www.iitk.ac.in/nerd/web/articles/adaptive-traffic-light-timer-
control-atltc/#.X9mHYWgzaUk
[6] Viola, Paul, and Michael Jones. "Robust real-time object detection."
International journal of computer vision 4.34-47 (2001): 4.
[7] Zhang Lei, Zhang Xue-fei and Liu Yin-ping. “Research of the Realtime Detection of
Traffic Flow Based on OpenCV”, in Proceedings of International Conference on Computer
Science and Software
Engineering, Hubei, 2008, vol.2, pp: 870 – 873.
Rainer Lienhart, Alexander Kuranov and Vadim Pisarevsky, “Empirical Analysis of
Detection Cascades of Boosted Classifiers for Rapid Object Detection”, Pattern Recognition,
Springer, vol. 2781, 2002, pp. 297-304.
[9] R.Chauhan and S.Bhadoria, “An Improved Image Contrast Enhancement Based on
Histogram Equalization and Brightness Preserving Weight Clustering Histogram
Equalization”, In Proceedings of IEEE International Conference on Communication Systems
and Network Technologies, Katra, Jammu,2011, pp. 597 – 600.
[10] Ali, Ihtisham & Malik, Arsalan & Ahmed, Waqas & Khan, Sheraz. (2015). Real-
Time Vehicle Recognition and Improved Traffic Congestion Resolution.
10.1109/FIT.2015.48.
[11] Rainer Lienhart, Alexander Kuranov and Vadim Pisarevsky, “Empirical Analysis of
Detection Cascades of Boosted Classifiers for Rapid Object Detection”, Pattern Recognition,
Springer, vol. 2781, 2002, pp. 297-304.

Dept of ISE 2022-23 17 | P a g e