Traffic Signal Violation Detection System
Supervised By
Mahmudul Hasan Shauqi
Lecturer
Dept. of Computer Science and Engineering
Presented By
Abu Noman Md. Sakib
Roll: 1607058
Pias Roy
Roll: 1607071
Department of Computer Science and Engineering
Khulna University of Engineering & Technology, Khulna
Acknowledgement
With immense please, We, Abu Noman Md. Sakib and Pias Roy presenting
project report of “Traffic Signal Violation Detection System” as part of the curriculum
of “CSE 3200: System Development Project”. We wish to thank all the people who gave
us unending support.
We express our profound thanks to our supervisor Mahmudul Hasan Shauqi sir who
have directly guided us and helped us in the development of this system project.
Abu Noman Md. Sakib
Roll: 1607058
&
Pias Roy
Roll: 1607071
Page | 2
TABLE OF CONTENTS
Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2. Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
3. Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4. System Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
5. Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-7
6. Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8-12
7. Conclusion & Recommendation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
8. References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Page | 3
1 Introduction
The increasing number of cars in cities can cause high volume of traffic, and implies
that traffic violations become more critical nowadays in Bangladesh and also around the
world. This causes severe destruction of property and more accidents that may endanger the
lives of the people. To solve the alarming problem and prevent such unfathomable
consequences, traffic violation detection systems are needed. For which the system enforces
proper traffic regulations at all times, and apprehend those who does not comply. A traffic
violation detection system must be realized in real-time as the authorities track the roads all
the time. Hence, traffic enforcers will not only be at ease in implementing safe roads
accurately, but also efficiently; as the traffic detection system detects violations faster than
humans. This system can detect traffic light violation in real-time. A user friendly graphical
interface is associated with the system to make it simple for the user to operate the system,
monitor traffic and take action against the violations of traffic rules.
2 Objectives
The goal of the project is to automate the traffic signal violation detection system and
make it easy for the traffic police department to monitor the traffic and take action against the
violated vehicle owner in a fast and efficient way. Detecting and tracking the vehicle and their
activities accurately is the main priority of the system.
3 Related Work
Presented in [1] is the implementation of real-time traffic violation detection in a
monitoring stream which utilized simultaneous video stream from different cameras using
parallel computing techniques. Another approach of implementing real-time traffic violation
detection was seen in [2], as they used video-based traffic detection through an improved
background-updating algorithm, thereafter track the moving vehicles by feature based tracking
method.
This project is inspired by above project but it is implemented using a self-developed
approach. Conventionally vehicle detection is referred as an object detection problem. To
detect moving vehicle objects from the road, YOLOv3 model is used which uses Darknet-53.
After detecting vehicles, violation conditions are checked.
Page | 4
4 System Overview
Figure-1: Flow diagram of traffic signal violation detection system.
The System consists of two main components -
• Vehicle detection model
• A graphical user interface (GUI)
First the video footage from the road side is sent to the system. Vehicles are detected
from the footage. Tracking the activity of vehicles, system determines if there is any violation
or not. Figure 1 shows how the system works.
The Graphical User Interface (GUI) makes the system interactive for the user to use.
User can monitor the traffic footage and get the alert of violation with the detected bounding
box of vehicle. User can take further action using the GUI.
Page | 5
5 Methodology
5.1 Vehicle Classification
From the given video footage, moving objects are detected. An object detection model
YOLOv3 is used to classify those moving objects into respective classes. YOLOv3 is the third
object detection algorithm in YOLO (You Only Look Once) family. It improved the accuracy
with many tricks and is more capable of detecting objects. The classifier model is built with
Darknet-53 architecture. Table-1 shows how the neural network architecture is designed.
Features:
1. Bounding Box Predictions:
YOLOv3 is a single network the loss for objectiveness and classification needs to be
calculated separately but from the same network. YOLOv3 predicts the objectiveness score
using logistic regression where 1 means complete overlap of bounding box prior over the
ground truth object. It will predict only 1 bonding box prior for one ground truth object and any
error in this would incur for both classification as well as detection loss. There would also be
other bounding box priors which would have objectiveness score more than the threshold but
less than the best one. These errors will only incur for the detection loss and not for the
classification loss.
2. Class Prediction:
YOLOv3 uses independent logistic classifiers for each class instead of a regular softmax
layer. This is done to make the classification multi-label classification. Each box predicts the
classes the bounding box may contain using multilabel classification.
3. Predictions across scales:
To support detection a varying scales YOLOv3 predicts boxes at 3 different scales. Then
features are extracted from each scale by using a method similar to that of feature pyramid
networks. YOLOv3 gains the ability to better predict at varying scales using the above method.
The bounding box priors generated using dimension clusters are divided into 3 scales, so that
there are 3 bounding box priors per scale and thus total 9 bounding box priors.
4. Feature Extractor:
YOLOv3 uses a new network- Darknet-53. Darknet-53 has 53 convolutional layers, its
deeper than YOLOv2 and it also has residuals or shortcut connections. Its powerful than
Darknet -19 and more efficient than ResNet-101 or ResNet-152.
Page | 6
5.2 Violation Detection
The vehicles are detected using YOLOv3 model. After detecting the vehicles, violation
cases are checked. A traffic line is drawn over the road in the preview of the given video
footage by the user. The line specifies that the traffic light is red. Violation happens if any
vehicle crosses the traffic line in red state.
The detected objects have a green bounding box. If any vehicle passes the traffic light
in red state, violation happens. After detecting violation, the bounding box around the vehicle
becomes red.
Page | 7
6 Implementation
6.1 Computer Vision
OpenCV is an open source computer vision and machine learning software library
which is used in this project for image processing purpose. Tensorflow is used for
implementing the vehicle classifier with darknet-53.
6.2 Graphical User Interface (GUI)
The graphical user interface has all the options needed for the software. The software
serves administration and other debugging purposes. We don’t need to edit code for any
management. For example, if we need to open any video footage, we can do it with the
Open item (Figure-2).
Figure-2: Initial user interface view.
Page | 8
Primarily, for the start of the project usage, the administrator needs to open a video
footage using ‘Open’ item that can be found under ‘File’ (Figure-2). The administrator can
open any video footage from the storage files (Figure-3).
Figure-3: Opening a video footage from storage.
After opening a video footage from storage, the system will get a preview of the
footage. The preview contains a frame from the given video footage. The preview is used to
identify roads and draw a traffic line over the road. The traffic line drawn by administrator
will act as a traffic signal line. To enable the line drawing feature, we need to select ‘Region
of interest’ item from the ‘Analyze’ option (Figure-4). After that administrator will need to
select two points to draw a line that specifies traffic signal.
Page | 9
Figure-4: Region of Interest (Drawing signal line)
Selecting the region of interest will start violation detection system. The coordinates
of the line drawn will be shown on console (Figure-5). The violation detection system will
start immediately after the line is drawn. At first the weights will be loaded. Then the system
will detect objects and check for violations. The output will be shown frame by frame from
the GUI (Figure-6).
Figure-5: Line Coordinates (from console)
Page | 10
Figure-6: Final Output (on each frame)
The system will show output until the last frame of the footage. In background a
‘output.mp4’ will be generated. The file will be in ‘output’ folder of ‘Resources’. The process
will be immediately terminated by clicking ‘q’.
After processing a video footage, the administrator can add another video footage from
the initial file manager (Figure-2). If the work is complete the administrator can quit using
‘Exit’ item from File option.
Libraries used for Graphical User Interface:
• Tkinter
Page | 11
The overall flow of the software:
Figure-7: Overall flow of the software.
Page | 12
7 Conclusion & Recommendation
The designed algorithm was effectively able to detect the type of violation specified on
this project which are denying traffic signal. The convergence of detection for the traffic
violation mentioned is dissimilar, since it has a different threshold condition. The system
provides detection for traffic signal violation. Further, the system is able to process one data at
a time. Also, the program runtime is somewhat slow, and can be improved by using a computer
with high speed processor specifications or GPU.
Future research about the application of the designed algorithm for other advanced
image processing techniques. Since, this may improve the program runtime of the system by
neglecting other unnecessary steps done in a background difference method. A computer vision
algorithm may be done instead to provide more intelligence in the system. Our future plan is
to implement the number plate detection with OCR support to make this system more robust.
8 References
[1] G. Ou, Y. Gao and Y. Liu, "Real Time Vehicular Traffic Violation Detectionin 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".
Page | 13
