JOURNAL OF CRITICAL REVIEWS

ISSN- 2394-5125 VOL 7, ISSUE 13, 2020

DENSITY-BASED SMART TRAFFIC LIGHT TIME

PREDICTOR
L. Lakshmi1, Alluru Lakshmi Sai Prashanthi2, Bande Susmitha3, Gajula Niharikha4, CH. Srinivasasulu5

1
Professor, Dept of CSE, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India.
E-mail: [email protected]
2
Student, Dept of CSE, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India.
3
Student, Dept of CSE, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India.
4
Student, Dept of CSE, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India.
5
Professor, Dept of CSE, BVRIT Hyderabad College of Engineering for Women, Hyderabad, Telangana, India.

Received: 20.04.2020 Revised: 22.05.2020 Accepted: 17.06.2020

ABSTRACT: The traffic congestion problem in urban areas increasing every day, there is an urgent need to
introduce modern equipment and technologies to improve the traffic control system in almost all countries.
Existing methods such as human control or timer are proving to be less efficient to improve this crisis. Some
existing traffic control systems have been proposed by measuring vehicle density in real-time with modern
ordinal image processing techniques. These traffic flow control systems offer noteworthy improvements in
vehicle management, time of response, reliability, automation of traffic systems, and all over efficiency above
the existing systems. To overcome the disadvantages of the existing system, we need an efficient system for
effective traffic flow light control through image processing with machine learning. This system will recognize
the images of vehicles instead of using sensors integrated into a pathway to capture the images of vehicles. These
images will be analysed and processed using edge detection algorithms to detect vehicles with OpenCV,
rendering to traffic circumstances, then timer setting for traffic controlling light will be managed. The
experimental results revealed that our system outperforms when compared with the existing system.
KEYWORDS: Traffic light predictor, Edge detection, White point count, OpenCV, Preprocessing, Edge count,
Canny edge detection.
© 2020 by Advance Scientific Research. This is an open-access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/) DOI: http://dx.doi.org/10.31838/jcr.07.13.258

I. INTRODUCTION
As the inhabitants of modern cities increase with each passing day, as traffic increases, leading to overcrowding
problems. Traffic jams cause many serious challenges and problems in large and populous cities. Owing to the
traffic flow congestion, additional time is also wasted. This stable escalation in the count of vehicles on the roads
has strengthened the significance of effective traffic flow management for optimizing the use of the current
capacity of roads. Environmental protection and high petrol costs are also responsible for the motivation to
reduce traffic flow delays. Traffic accidents are another major problem in the contemporary world [1]. If we look
closely at the cause of traffic accidents, we have discovered that slender transportations and the hasty growth in
vehicles are the main causes for the increase in the number of traffic accidents.
When the traffic flow demand is high enough to interact among vehicles reduces the speed of movement as a
result, leads to some congestion. Like a demand approaching the capacity of the road (or junctions along the
way) extreme traffic jams form. Traffic congestion is defined as at a particular point when the vehicles are halted
for timeframes [1]. Traffic flow congestions can lead to drivers engaged and frustrated in the wrath of the road.
To avoid traffic jams, in traffic between, traffic Signal acknowledgment (TSA) is used to control traffic signals,
advice the driver and orders, or prohibits certain actions.
As the numbers of vehicles in big cities of every country are increasing every day and when traffic is increasing
day by day, there is a crucial requisite to a completely innovative traffic flow controlling system with innovative
technology by using already existing entire infrastructures [2]. To build new roads, spring roads, higher speed
lines, etc. need planning, big capital, and a lot of time; the focus should be aimed at more efficient use of
prevailing infrastructure completely and effectively. Most of the existing methods are based on calculating the
vehicle count, a total pixel count of the vehicles.

1627
 JOURNAL OF CRITICAL REVIEWS
ISSN- 2394-5125 VOL 7, ISSUE 13, 2020

But, evaluation of the count of the number of automobiles, can provide incorrect results when the vehicle area is
very trivial (vehicles can be found very near with one another can be calculated to one). It is unable to count
auto-rickshaws and rickshaws as vehicles, which are the main source of traffic flow particularly in Asian
countries. Counting the pixels has the drawback of counting light things as vehicles for example pedestrians and
footpaths are also taken into count. Most of the existing works mainly concentrated on the density of traffic [3].
But this may give fruitful results because the lanes where the frequency is less and wasting a lot of time at a
signal point.
The current traditional system of traffic lights causing a tragic system of congestion, which is not a system of
traffic lights intelligent and reduce the definition of time and time and calculate the density of traffic and time.
The mechanism is different from the electrical sensors and the process. In this case, the system identifies vehicles
and produces vehicles for the function of time and density of vehicles. The overload is also produced by the
electronic control sensor. The use of electronic sensors is not effective because it has a high tendency to be
affected by cost and noise [1,4]. This is since there needs to be significant changes required in controlling the
traffic flow using traffic lights.
The proposed paper suggests a procedure in which the traffic mass is evaluated by associating the image
captured through live cameras with the traffic data in real-time concerning the images of the empty roads as a
reference image. Traffic flow light system through digital image processing consists of the system that can
recognize vehicles by images as an alternative of using sensors integrated with the pathway. The camera can be
used through traffic flow lights. It can seizure image sequences and perform the counting, correspondingly alters
the time allocated for a particular lane. It indicates that you can reduce traffic flow congestion by avoiding
wasting time with the green light preceding the empty path. It also produces more accurate results in
distinguishing the presence of the car, footpaths, and pedestrians, since it uses real images of traffic. It is mainly
based on real-time traffic content, so it outperforms than other systems that depend on the detection of the count
of the vehicle or pixel count.

II. LITERATURE REVIEW

The automatic traffic flow light is controlled by electric sensors and timers. In the customary traffic lights
system, all phases have a constant number of loaded loads in the timer. The lights are automatically and
mechanically controlled by the change in the timer value.
The expansion of VANET networks (ad hoc vehicles), which are an important part of the new categories of
networks developing in wireless technology. The characteristics of VANET are to deliver communication among
the vehicles themselves and between the vehicles and the roadside units. VANETs also play a significant role in
concepts like smart cities. This work is based on a smart city framework that will convey traffic information and
go a long way in helping drivers make smart decisions and extemporaneous to avoid traffic jams, which will
eventually help to reduce overall traffic jams [5]. This system undergoes from the limitation that the
implementation of VANET requires the installation of suitable hardware in each vehicle, which can be relatively
difficult to install in a two-wheeled vehicle. Besides, the entire outline is user-dependent, as overall traffic flow
congestion will depend on the decision made by the user.
The edge detection technique is necessary to extract the necessary traffic information from the footage of CCTV.
It is used to segregate the necessary information as of the rest of the image. Various edge detection techniques
are available and they may vary mainly in terms of their characteristics such as detection sensitivity, noise
reduction, and precision [6]. The Canny edge detection technique has been observed to show higher accuracy in
detecting an object with peak signal to noise ratio, higher entropy, runtime and root mean square error compared
to other techniques.
Another model for a smart traffic system that will summarize the surveillance features through the cameras
present at the intersection, and with the assistance of data transmission systems will allow users to have the right
to use this data. Background/ foreground modeling and image analysis schemes are essentials of monitoring and
data transmission over an ad-hoc mobile network that will include a system-wide data transmission part [7].
Several experiments have been carried out in the project, which shows great potential in terms of real-time
performance and efficiency.
An intelligent integrated traffic light control and management system with photoelectric sensors in which
connection of the sensors is a most important criterion in the system since the traffic flow control department
have to observe the cars that move with certain traffic and then handover this data to the traffic flow control

1628
 JOURNAL OF CRITICAL REVIEWS
ISSN- 2394-5125 VOL 7, ISSUE 13, 2020

department, which will control the traffic lights according to the sensor readings using an algorithm based on the
comparative density of all roads [1, 8].
A smart traffic flow control system based on Nios-II and DSP. Its smart traffic flow management system model
deploys a dual-processor combined with Field Programmable Gate Array Logic Control (FPGA), which includes
features such as interface configuration, sharing, and establishment of related information, and human-computer
interaction [7]. To achieve congestion in the car, it differs from the conventional traffic flow signal controller in
that it operates mainly in synchronization mode and multiple phases according to the dynamic requirements of
the user.
To overcome the above disruptions, a system is proposed in which traffic flow density is measured by comparing
images captured through CCTV with real-time traffic information relative to the empty road image as a reference
image. Techniques such as the use of aerial imaging, complex sensor-based systems, and the use of additional
devices are more advantageous and the grayscale image has an excellent signal/noise ratio compared to an RGB
image.

III. THE PROPOSED SYSTEM ARCHITECTURE

Figure 1: Real Time Traffic Light Allocation System

The architecture of the proposed system is shown in Figure 1. Initially the real-time traffic images are captured
through CCTVs present at the traffic light system junctions. The real-time traffic image will be uploaded to the
application. The application process the image mainly in four modules. The first module performs the
preprocessing of the image, then we apply the canny edge detection algorithm to perform the edge detection, as a
next step we are evaluating the white pixel count in the real-time image. Finally we will calculate the average of
original and reference images white pixel count, and then based on white pixel count then we will allocate the
timer for the green light.

1. Preprocessing of the Traffic Image

The real-time image obtained through CCTV is an RGB image that is colored. During the preprocessing the
image is to be converted into a gray-scale image. As a first step the image is to be converted into matrix form
using the Numpy arrays and then perform the rotation of the matrix. Next we performed the cropping of the
image to reduce the size of the image. Then we have used Gaussian blur to function to blur the image.

1629
 JOURNAL OF CRITICAL REVIEWS
ISSN- 2394-5125 VOL 7, ISSUE 13, 2020

2. Edge Detection Using Canny Edge Detection

It is one of the efficient edge detection algorithms; by using this algorithm first we have removed noise from the
image using a Gaussian filter. Next we have applied Sobel-filter to flatten the image in both vertical and
horizontal directions, and then from this find we have calculated the gradient of edges and each pixel direction.
After this we have performed edge suppression that is the removal of pixels that will not contribute to the
formation of edges. As the last step we have performed the Hysteresis threshold to determine which edges are to
be considered as edges and which are not to be considered for this which has defined low threshold and high
threshold, if our edges lie above low and high threshold values are considered to be edges. The edges which lie
below low threshold values are to be neglected.

3. White Pixel Count Calculation of the Image

After applying the canny edge detection algorithm the resulting image consists of the edged of all vehicles in
white color. In this step we have calculated the white pixel count of the original image and reference image using
the NumPy array with image sum function. Here we have used image value as 255 to count only white pixels.

4. Green Light Signal Time Allocation Calculation

After calculating white pixel count in the reference image and original image then we have average count. Based
on the average count we have estimated time allocation for the green light signal.
Average count = (Original image white pixel count)/(Reference image white pixel count)*100
After calculating the average count if the average count greater than 90 then the green light timer is set to 120
due to high traffic conditions. The average count lies between 70 and 90 then the timer is set to 60 seconds due to
moderate conditions. The average count lies between 50 and 70 then the timer is set 40 seconds due to less
moderate traffic conditions. The average count is less than 50 then the timer is set to 20 seconds due to very little
traffic.
In this paper, we compare the performance of SCMA, MUSA, and PDMA for different users such as 6, 15 and
22. The SCMA provides better SER for large number of users because of sparse coding used in LDPC decoding
schemes. The Symbol error rate performance is calculated and plotted for different number of users, the analysis
of PDMA and SCMA is examined for same factor graph. OFDM scheme is used as transmitter with each
subcarrier is connected to mapping scheme of NOMA techniques. The average SER is calculated for three
NOMA techniques for different users in the hexagonal shape cell. The SCMA shows 3% improvement in
average SER value when compare to PDMA and MUSA techniques.

IV. RESULTS AND DISCUSSIONS

We have evaluated the performance of our system with high, average, and low traffic conditions and the results
revealed that our system performance is good when compared with existing systems. The Figure 2 shown below
represents the settings of the green signal light timer in the high traffic condition.

1630
 JOURNAL OF CRITICAL REVIEWS
ISSN- 2394-5125 VOL 7, ISSUE 13, 2020

Figure 2: A Real Time Image with Heavy Traffic

A real-time image as shown in Figure 2 is uploaded to the system to analyze and find the green signal time
allocation.

Figure 3: A Preprocessed Image

The preprocessed image after applying the canny edge detection algorithm is shown in Figure 3.

Figure 4: White Point Count Calculation

From the gray scale image we have calculated white point count which is shown in Figure 4.

1631
 JOURNAL OF CRITICAL REVIEWS
ISSN- 2394-5125 VOL 7, ISSUE 13, 2020

Figure 5: Green Signal Time Calculation

The given original image contains more vehicles so 120 seconds is allocated as shown in Figure 5 to for the
green light in that lane.
Similarly we evaluated the performance of our algorithm for fewer traffic conditions.

Figure 6: A Real Time Image with Less Traffic

A real-time image as shown in Figure 6 is uploaded to the system to analyze and find the green signal time
allocation.

Figure 7: A Preprocessed Image

The preprocessed image after applying the canny edge detection algorithm is shown in Figure 7.

1632
 JOURNAL OF CRITICAL REVIEWS
ISSN- 2394-5125 VOL 7, ISSUE 13, 2020

Figure 8: White Point Count Calculation

From the grayscale image we have calculated white point count which is shown in Figure 8.

Figure 9: Green Signal Time Calculation

The given original image contains a very less number of vehicles so 20 seconds is allocated as shown in Figure 5
for the green light in that lane.

V. CONCLUSION
In the proposed paper we have recommended an innovative solution for the traffic control system in urban cities
where the traffic conditions are merely the most horrible day by day. In the proposed method we have used real-
time traffic images instead of sensors. These images are preprocessed using machine learning image
preprocessing techniques and performed optimized edge detection using a canny edge detection algorithm. After
obtaining preprocessed grayscale images then we have performed white point count in both images captured in
real-time and reference image. Then we have calculated average count based on average count we have
performed green signal time allocation. The results revealed that the performance of our system is good when
compared with traditional and semi-automated systems.

VI. REFERENCES
[1] Taqi Tahmid, and Eklas Hossain. "Density based smart traffic control system using canny edge
detection algorithm for congregating traffic information", 3rd International Conference on Electrical
Information and Communication Technology (EICT), 2017.
[2] John, Vijay, Liu, Zheng , Yoneda, Keisuke, Mita, S & Qi, “ Traffic Light Recognition in Varying
Illumination using Deep Learning and Saliency Map” in International conference on intelligent
transportation system, 2014.
[3] T. Semertzidis, K. Dimitropoulos, A. Koutsia and N. Grammalidis, “Video sensor network for real-time
traffic monitoring and surveillance,” in The Institution of Engineering and Technology, Volume 4(2),
2010, pp. 103 – 112
[4] Zhao, Biao, Zhang, Chi, Zhang, and Lichen, “Real-Time Traffic Light Scheduling Algorithm Based on
Genetic Algorithm and Machine Learning”, in an Approach for Map-Matching Strategy of GPS-
Trajectories Based on the Locality of Road Networks. 385-398,2015.

1633
 JOURNAL OF CRITICAL REVIEWS
ISSN- 2394-5125 VOL 7, ISSUE 13, 2020

[5] G.T. Shrivakshan and Dr.C. Chandrasekar, “A Comparison of various Edge Detection Techniques used
in Image Processing,” in international Journal of Computer Science Issues, Vol. 9, Issue 5, No 1,
September,2012.
[6] Lakshmi, L.,Bhaskara Reddy, P., ShobaBindu, C. and Aravind Kumar, N. (2019) ‘Dynamic navigation
of web query results using B-tree and improved page rank algorithm’, Int. J. Computing Science and
Mathematics, Vol. 10, No. 2, pp.193–202.
[7] Jiaqiao He, LiyaHou and Weiyi Zhang, “A Kind of Fast Image Edge Detection Algorithm Based on
Dynamic Threshold Value,” Sensors & Transducers, Vol. 24, Special Issue, August 2013, pp. 179-183.
[8] S. Sobitha Ahila, D. Sivakumar M. Naveen Nanda, Nivedita, Suresh Kumar Nair, SruthiKannan,
“Traffic Lights Analysis and Manipulation for Emergency Vehicles Using Deep Learning Algorithm”,
in International Journal of Recent Technology and Engineering (IJRTE), 7(6S5), 2019.

1634