e-ISSN: 2582-5208

International Research Journal of Modernization in Engineering Technology and Science

( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:02/February-2022 Impact Factor- 6.752 www.irjmets.com

TRAFFIC RULES VIOLATION DETECTION SYSTEM

Mr. M. V. Rajesh*1, Mr. T. Soma Sekhar*2, Ms. P. Mounika Devi*3,
Ms. Y. Mounica Bhavani*4, Mr. R. S. S. V. S. S. Narendra*5, Mr. K. Bulli Raju*6
*1Associate Professor, Department Of CSE, Pragati Engineering College, Surampalem, A.P, India.
*2Assistant Professor, Department Of CSE, Pragati Engineering College, Surampalem, A.P, India.
*3,4,5,6B.Tech. III Year Student, Department Of CSE, Pragati Engineering College,
Surampalem, A.P, India.
ABSTRACT
Nowadays, there is rapid growth in the population and people are spending too much for their comforts and
luxuries. As a result, more automobiles are being purchased, particularly in urban areas. This can result in
severe traffic, indicating that traffic violations are becoming more hazardous in every place of the world. As a
result, people's awareness decreases, and there are more accidents, which may result in the loss of many of
lives. These situations make us recognize the need to develop traffic violation detection systems to automate
traffic regulations and make people aware of it. The proposed traffic violation system can identify signal
violations, and the individuals are informed that they will be recognized if they break a traffic law.
Keywords: Automation, Detection, Regulation, Signal, Traffic, Violation.
I. INTRODUCTION
The growing number of vehicles in all crowded cities can lead to large amount of traffic which indicates that
traffic violations become more dangerous all around the world. This results in the drastic destruction of
property and accidents that may threaten the people. To solve the disturbing problem and prevent such
immeasurable consequences, traffic violation detection systems are required. To do this, the system invokes
appropriate traffic regulations and punish people who don't follow. A detection system should be used for
violation detection in real-time as the officials trace the roads and vehicles. So the traffic regulators can use this
easily to maintain safe roads efficiently, as the system identifies violations with a higher speed than people. A
user-friendly GUI is combined with the traffic violation system to make it easy for the users to deal with the
system, track the traffic and take necessary actions against the traffic violations. This is able to identify the most
usual kinds of violation. Detecting and keeping track of the vehicles and their activities effectively is the first
priority of this system.
II. METHODOLOGY
The proposed system can be implemented as follows:
1. Vehicle Classification
A video footage is given as input to the system and the moving objects are detected. A detection model YOLO
version3 is used to categorize the moving vehicles into particular classes. YOLOv3 is the third version algorithm
in YOLO family which is used for detection of objects. It maintains the integrity of data with the use of several
techniques and it is able to identify the objects more accurately. Darknet-53 architecture is used to build the
classifier model.
Features of vehicle classification are:
 Bounding Box Predictions
 Class Prediction
 Predictions across scales
 Feature Extractor
Bounding Box Predictions:
It is a single network for real-time object detection and categorization needs to be evaluated individually but
from the same convolutional region. This algorithm anticipates the fair score by means of logistic regression.
Here 1 indicates the overall overlapping of bounding box on the object. It predicts only 1 bounding box prior for
one ground truth object and any misconception in this process lead to both categorization and identification
www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science
[68]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:02/February-2022 Impact Factor- 6.752 www.irjmets.com
loss. There will be some more bounding box priors whose value may be greater than the threshold limit and
less than the best value. This type of errors will only incur for the identification loss but not for the
classification loss.
Class Prediction:
This algorithm makes use of unconventional logistic classifiers for every class in place of a conventional
Softmax layer. This process is performed to convert into a multi-label classification. Every box identifies the
classes that it may contain with the help of the multilabel classification.
Predictions across scales:
For the identification of different scales, it detects boxes at three different scales. Then features can be extracted
from each scale by using a method which is similar to that of feature pyramid networks. YOLOv3 gains the
ability to predict at different scales using the above method. The bounding box priors which are generated
using dimension clusters can be divided into 3 scales, so that there are three bounding box priors per scale.
Thus there will be 9 bounding box priors as a total.
Feature Extractor:
It uses a new network called Darknet-53 which consists of 53 convolutional layers, and it has more features
when compared to YOLO version2. It is more powerful than Darknet -19. And its efficiency is higher than
ResNet-101/ResNet-152.
2. Violation Detection
The vehicles are detected using the model YOLOv3. After detection of vehicles, violation cases are checked. A
traffic line is drawn by the user over the road in the preview of the given video footage. This line indicates that
the traffic signal is red in that area. The objects are detected with a bounding box of green color around them. If
any of the moving vehicles go ahead of the traffic line indicating the red signal, then the traffic violation
happens. The color of the bounding box around the vehicle becomes red after the violation detection.
III. TECHNOLOGIES
COMPUTER VISION
OpenCV is a field of AI that helps systems and computers to obtain meaningful and useful information from
images, visual inputs and videos. OpenCV is an open-source computer vision and Python library which is
generally used for the purpose of image processing. TensorFlow is used for classification of the vehicles with
the help of darknet-53.
GRAPHICAL USER INTERFACE
The GUI contains all the options required for the maintenance of the software. This can be used for
management and other troubleshooting objectives. The users do not need to edit the code for any other
management purposes. Tkinter library can be used to build the interface. Tkinter is the standard library in
Python for building GUI. Python in association with this library provides a very fast and simple way to build
applications based on GUI.

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[69]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:02/February-2022 Impact Factor- 6.752 www.irjmets.com
IV. SYSTEM DESIGN

Figure 1: Architecture of Violation Detection System

This system is made up of two major elements -
 Vehicle Detection Model
 Graphical User Interface
Firstly, a video footage of traffic is given to the system as input. Vehicles will be identified from the given video.
By tracing the activities of the moving vehicles, the system determines if any violation happened or not. User
can keep track of the traffic in the footage and be notified of violation happened with the bounding box of the
vehicle identified. After the user has got the alert, he can take the necessary actions using the GUI.
V. RESULTS AND DISCUSSION

Figure2: Home Page

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science
[70]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:02/February-2022 Impact Factor- 6.752 www.irjmets.com

Figure3: Opening input video

Figure 4: Preview loaded for specific video

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[71]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:02/February-2022 Impact Factor- 6.752 www.irjmets.com

Figure 5: Region of Interest

Figure 6: Drawing signal line

Figure 7: Vehicle Classification

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[72]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:02/February-2022 Impact Factor- 6.752 www.irjmets.com

Figure 8: Violation Detection

Firstly, for beginning the project, the user requires to open a video using “Open” option that is visible under File
(Figure-2). The user can open any footage from the system storage (Figure-3). After opening it, the system will
load a preview (Figure-4). It contains one of the frames of the input video. It is used to recognize the roads and
select region of interest. The selection by the user will act as a traffic line. To draw this line, we need to click on
“Region of interest” item from the Analyze option (Figure-5). After clicking on it, the user needs to pick two
points to plot a line that indicates the traffic signal (Figure-6). Selection of the traffic line will automate the
process of violation detection. Now the system identifies the vehicles and then it is ready to detect the
violations. After that, GUI gives output frame by frame (Figure-8). The system continues to display the result
until it reaches the final frame of the video footage. In the background, an output file will be generated. This file
can be found in output folder of Resources. This process of violation detection will be terminated by clicking q.
After applying the process to all the frames of the input video footage, the user can give another video as input
(Figure-2). If the work is completed the user can quit using Exit item available in the File option.
VI. CONCLUSION
If a vehicle goes ahead of the predefined traffic line on the road, it is recognized as a traffic violation. The
proposed system is faster and works more efficiently than human. As we all know, traffic police is the one who
captures the image of individuals violating traffic regulations but the traffic police won’t be capable of detecting
and capturing more than one violation simultaneously. The algorithm used here was effectively able to identify
the kind of violation described here which is violating the traffic signal. The current system provides detection
for traffic signal violation. Further, the system will be capable of processing a single data at a time. And also, the
program has slower runtime, which can be further enhanced by using a system containing high processing
speed. Further research is required for applying the existing algorithm for other high level image processing
techniques as it improves the system’s runtime by ignoring other unnecessary things. An OpenCV algorithm can
be used in place of the existing one to enhance the performance of the system we are using now. Our further
aim is to detect the number plate of the vehicle that violates the traffic signal and add more traffic violation
conditions to improve the efficiency of the system.
ACKNOWLEDGEMENT
“We express our sincere gratitude to our and learned and respected guides, Associate Professor, Mr. M. V.
Rajesh and Assistant Professor, Mr. T. Soma Sekhar for their valuable help and guidance. We are very thankful
to them for the encouragement they had given us in completing this project. We are also grateful to respected
Prof. RADHIKA MANI HOD (CSE Dept.) and to our respected Director for permitting us to utilize all the
necessary facilities of the institution. We are also thankful to all the other faculty & staff members of our
department for their help and cooperation. Lastly, we would like to express our appreciation towards our

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[73]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:04/Issue:02/February-2022 Impact Factor- 6.752 www.irjmets.com
classmates and friends and our indebtedness to our parents for providing us the moral support and
encouragement while doing this project.”
P. Mounika Devi –19A31A0525
Y. Mounica Bhavani – 19A31A0543
R.S.S.V.S.S Narendra- 19A31A0558
K. Bulli Raju -20A35A0505
VII. REFERENCES
[1] G. Ou. Y. Gao and Y. Liu, "Real Time Vehicular Traffic Violation Detection in Traffic Monitoring System,"
in 2012 IEEE/WIC/ACM, Beijing, China , 2012.
[2] X. Wang, L.-M. Meng, B. Zhang, J. Lu and K.-L. Du, "A Video-based Traffic Violation Detection System," in
MEC, Shenyang, China, 2013.
[3] Joseph Redmon and Ali Farhadi, "YOLOv3: An Incremental Improvement".
[4] Mukremin Ozkul, Ilir Capuni (2018).” Police-less multiparty traffic violation detection and reporting
system with privacy preservation”, IET Intelligent Transport Systems, Vol. 12 No. 5, pp. 351-358.
[5] Wang J., Chen, Z., Li P., Sheng B., & Chen R (2019), ‘Real Time Non-Motor Vehicle Violation Detection in
Traffic Scenes’ 2019 IEEE International Conference on Industrial Cyber Physical Systems (ICPS), pp.
724728.

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[74]