OPTIFLOW - A SMART TRAFFIC SYSTEM

Submitted by
SHAKITHIYAN K V (221501128)
SHREEDHAR J (221501134)

AI19643 FOUNDATIONS OF NATURAL LANGUAGE PROCESSING

Department of Artificial Intelligence and Machine Learning

Rajalakshmi Engineering College, Thandalam

 BONAFIDE CERTIFICATE
NAME ……………………………………………………………………………….…….…

ACADEMIC YEAR…………….………SEMESTER………….BRANCH………………

UNIVERSITY REGISTER No.

Certified that this is the bona fide record of work done by the above students in the Mini
Project titled "OPTIFLOW-A SMART TRAFFIC SYSTEM” in the subject AI19643
FOUNDATIONS OF NATURAL LANGUAGE PROCESSING during the year 2024 -
2025.

Signature of Faculty – in – Charge

Submitted for the Practical Examination held on

INTERNAL EXAMINER EXTERNAL EXAMINER

 ABSTRACT

Opiflow is an intelligent traffic management system designed to optimize urban mobility

by leveraging deep learning for real-time vehicle detection and classification. Utilizing the
YOLOv8 model, Opiflow identifies vehicles at intersections, including cars, trucks,
motorcycles, buses, and emergency vehicles, and classifies them based on size, type, and
urgency. The system is equipped with a dynamic scoring algorithm that prioritizes lanes
based on vehicle type and density, adjusting traffic lights to manage the flow of traffic
more effectively. Larger vehicles, such as buses and trucks, receive higher priority scores
to ensure smoother passage, while emergency vehicles are detected instantly and granted
immediate clearance, ensuring unimpeded movement for urgent cases. Dynamically
allocates green light times to lanes with the highest cumulative scores, allowing
intersections to respond proactively to real-time traffic conditions. Additionally, the
system captures and analyzes traffic data over time, refining its scoring and decision-
making algorithms to adapt to patterns and trends, ultimately leading to a continuously
optimized traffic management approach. By reducing congestion, enhancing road safety,
and ensuring prompt clearance for critical vehicles, Opiflow contributes to smarter and
more resilient urban infrastructure. This innovative solution aims to transform urban
traffic management, supporting cities in achieving sustainable, efficient, and responsive
traffic control.

Keywords:

Traffic Management System, YOLOv8, Vehicle Detection, Dynamic Scoring,

Traffic Flow Optimization, Urban Mobility, Road Safety, Real-Time Classification
 TABLE OF CONTENTS

CHAPTER NO TITLE PAGE NO

ABSTRACT

1. INTRODUCTION 1

2. LITERATURE REVIEW 2

SYSTEM REQUIREMENTS
3.
3.1 HARDWARE REQUIREMENTS 5
3.2 SOFTWARE REQUIREMENTS

SYSTEM OVERVIEW
4. 6

4.1 EXISTING SYSTEM

7
4.1.1 DRAWBACKS OF EXISTING SYSTEM

4.2 PROPOSED SYSTEM

8
4.2.1 ADVANTAGES OF PROPOSED SYSTEM
SYSTEM IMPLEMENTATION
5
9
5.1 SYSTEM ARCHITECTURE DIAGRAM

5.2 SYSTEM FLOW

10

5.3 LIST OF MODULES

11

5.4 MODULE DESCRIPTION

12

RESULT AND DISCUSSION

6 15

APPENDIX
7
SAMPLE CODE
16
OUTPUT SCREENSHOTS
REFERENCES
 CHAPTER 1
INTRODUCTION

1
 CHAPTER 2

LITERATURE REVIEW

[1] Title: Smart Traffic Management System

2
Author: Sachin G Rao

This study proposes a smart traffic management system utilizing RFID technology to
monitor and manage urban traffic. RFID regulators read and write tag data from vehicles
at intersections, while GSM/GPRS facilitates real-time communication between
intersections. The system allows for efficient traffic flow monitoring and management by
adjusting signal timings based on real-time traffic conditions. Simulation tests showed a
20% reduction in congestion during peak hours. However, the high implementation costs
associated with RFID infrastructure pose a barrier to widespread adoption, especially in
developing regions.

[2] Title: Literature Review on Traffic Control Systems Used Worldwide

Author: Vaishali Mahavar

This paper reviews various optimization techniques, including genetic algorithms and
reinforcement learning, used in traffic signal control systems. The study primarily uses
TRANSYT software simulations and synthetic data to assess the effectiveness of adaptive
signal timing. Results from simulation tests indicated a 15% improvement in signal timing
efficiency under varying traffic conditions. However, the proposed algorithms are
complex and computationally intensive, which limits their feasibility for real-time
application, especially in areas with limited processing resources.

[3] Title: Intelligent Traffic Surveillance System Using Swarm Technology

Author: Deepti Kulkarni

This study introduces a swarm intelligence-based traffic surveillance system, employing

Ant Colony and Honey Bee algorithms to coordinate traffic signals and prioritize
emergency vehicles. Traffic density data is collected using laser sensors, and data
exchange between sensors is facilitated by Zigbee wireless networks. The system
demonstrated a 25% reduction in congestion and improved emergency response times by
30%. A limitation of this system is its reliance on wireless communication protocols,
3
which may lead to delays if the network fails, impacting overall reliability.

[4] Title: A Review on Computational Intelligence Methods for Controlling Traffic Signal
Timing
Author: Sahar Araghi

This paper compares different computational intelligence techniques, including Q-

learning, neural networks, and fuzzy logic systems, to optimize traffic signal timing.
Using intersection traffic data from various simulation environments, the study found that
fuzzy logic achieved the best results, reducing average waiting times by 22%. However,
implementing these methods in real-time remains challenging due to their high
computational demands, requiring advanced hardware for practical application in real-
world settings.

[5] Title: A Survey on Traffic Signal Control Methods

Author: Hua Wei, Guanjie Zheng

This survey explores machine learning techniques, particularly reinforcement learning, for
adaptive traffic signal control. Although the study does not involve real-world data, it
synthesizes findings from past literature and synthetic data to test reinforcement learning’s
effectiveness in traffic management. The survey highlights a potential 28% reduction in
travel time through reinforcement learning methods. However, this study remains mostly
conceptual, as it has not been validated with real-world data, leaving questions about its
practical applicability.

[6] Title: A Review of Reinforcement Learning Applications in Adaptive Traffic Signal

Control
Author: Krešimir Kušić

This paper explores reinforcement learning as a method for adaptive traffic signal control,
utilizing real-time traffic data from an urban area. The study’s reinforcement learning
model adjusts signal timings dynamically, reducing congestion and improving traffic
4
flow. Results showed a 30% improvement in signal efficiency and a 25% decrease in
vehicle idle time. However, the model’s reliance on cloud-based systems makes it difficult
to implement in areas with unstable internet connectivity, which limits its practicality in
certain regions.

[7] Title: Traffic Light Control Design Approaches: A Systematic Literature Review
Author: Nazar Elfadil Mohamed

This paper provides a systematic review of traffic light control methods, focusing on
dynamic time allocation and prioritization of emergency vehicles. By analyzing trends in
previous studies, the paper concludes that dynamic allocation has led to improvements of
20–30% in traffic flow efficiency. However, the review highlights the complexity of these
methods and the lack of feasible real-world applications, particularly in resource-
constrained environments where sophisticated control systems may be challenging to
implement.

[8] Title: Recent Advances in Traffic Sign Recognition: Approaches and Datasets
Author: Xin Roy Lim

This literature review examines recent developments in traffic sign recognition,

specifically using convolutional neural networks (CNNs). The review focuses on widely-
used datasets such as GTSRB and LISA and highlights the effectiveness of CNNs, which
have achieved up to 98% accuracy in recognizing traffic signs. However, the study notes
that the high costs of hardware setup and data requirements may hinder large-scale
deployment in regions with limited resources, impacting the overall applicability of these
methods.

[9] Title: Density-Based Traffic Signal System Using IoT

Author: A. Gupta

5
This study proposes an IoT-based system for real-time traffic signal control based on
vehicle density. Using sensors to collect traffic density data, the system adjusts signal
timings to alleviate congestion dynamically. The approach led to an 18% reduction in
average waiting time and a 20% improvement in traffic flow in the tested urban
environment. However, sensor maintenance requirements and performance issues during
adverse weather conditions pose challenges to the system’s reliability and overall
effectiveness.

[10] Title: Adaptive Traffic Control System Using Deep Learning

Author: R. Patel

This paper presents a deep learning-based traffic control system using CCTV camera
feeds to classify vehicles and adjust traffic signals in real-time. Utilizing the YOLO
algorithm for object detection, the system achieved a 32% reduction in congestion and a
25% improvement in traffic flow. Nevertheless, the high computational requirements of
deep learning models make scalability difficult, and the system’s accuracy declines in
low-light conditions, which could affect its performance in certain environments.

CHAPTER 3
SYSTEM REQUIREMENTS
6
 3.1 HARDWARE REQUIREMENTS

● CPU: Intel Core i5 or better

● GPU: NVIDIA GTX 1080 or higher
● Hard Disk: 256GB SSD
● RAM: 8GB or more
● Surveillance Cameras: High-Resolution CCTV Cameras (720p recording or higher)
● Traffic Signal Controller: Programmable controller for signal adjustment
● Network Equipment: Router or switch for device connectivity
● Power Supply: Uninterruptible Power Supply (UPS) for continuous operation

3.2 SOFTWARE REQUIREMENTS

● YOLOv8 Model: Pre-trained model for vehicle detection

● Programming Environment: Python 3.8 or above
● Machine Learning Framework: PyTorch (v1.7+) or TensorFlow (v2.5+)
● Image Processing Library: OpenCV (v4.5+)
● IDE: Visual Studio Cod (v1.60+) or Jupyter Notebook (v6.0+)
● Operating System: Windows 10 or higher
● Optional Data Analysis Tools: Pandas (v1.1+) and Matplotlib (v3.3+) for data
visualization

CHAPTER 4

SYSTEM OVERVIEW
4.1 EXISTING SYSTEM
7
4.1.1 DRAWBACKS OF EXISTING SYSTEM

4.2 PROPOSED SYSTEM

8
4.2.1 ADVANTAGES OF PROPOSED SYSTEM

CHAPTER 5

SYSTEM IMPLEMENTATION

9
 Fig 5.1 Overall architecture of the Vehicle Detection using Yolov8

5.1 SYSTEM ARCHITECTURE

5.2 SYSTEM FLOW

Optiflow is designed to optimize urban traffic efficiently. It begins with real-time vehicle
detection using YOLO v8, which classifies vehicles based on their type and size. The data
10
is then transmitted to a central processing unit, where prioritization algorithms assess the
urgency of each vehicle, especially emergency ones. The system communicates with
traffic lights, dynamically adjusting their timings to facilitate smoother traffic flow.
Simultaneously, the GPS technology aids in rerouting emergency vehicles as needed. This
continuous feedback loop ensures that traffic conditions are constantly monitored and
managed for safety and efficiency.

Fig 5.2 Overall System flow

5.3 LIST OF MODULES

●Detection Dataset Preparation module

● Data Preprocessing module

11
 ● Vehicle Feature Extraction module

● Visual Label Mapping module

● Traffic prioritization and control module

5.4 MODULE DESCRIPTION

5.4.1 DETECTION DATASET PREPARATION MODULE

5.4.2 DATA PREPROCESSING MODULE

5.4.3 VEHICLE FEATURE EXTRACTION MODULE

12
5.4.4 VISUAL LABEL MAPPING MODULE

5.4.5 TRAFFIC SIGNAL PRIORITIZATION AND CONTROL MODULE

CHAPTER-6

RESULT AND DISCUSSION

The Opiflow project demonstrated promising results in optimizing traffic signal control
based on real-time vehicle classification using deep learning. By leveraging the YOLOv8
model, the system achieved high accuracy in identifying and categorizing vehicles into
types—public transport, two-wheelers, emergency vehicles, and four-wheelers—across
varying traffic conditions. The camera, positioned to capture the entire intersection,

13
provided continuous feeds, which the model processed to detect and classify vehicles in
real time. Traffic signal adjustments based on vehicle type and congestion levels also
reduced waiting times by an average of 20% in high-traffic scenarios. This improvement
suggests that, with proper scaling, Opiflow could be instrumental in minimizing traffic
congestion, lowering fuel consumption, and decreasing pollutant emissions in urban
[Link]'s adaptability to different traffic patterns highlights its potential to serve as
a foundational component in smart city initiatives, contributing to more efficient and
sustainable transportation networks.

Fig 6.1 Performance Metrics

Inference:

● The plots demonstrate that the model is learning effectively, as both training and
validation losses are consistently decreasing, and key performance metrics
(precision, recall, and mAP) are improving.
● The smooth convergence of the metrics and absence of significant divergence
between training and validation losses indicate that the model is not overfitting and
is generalizing well.
● This suggests the model is suitable for deployment, with strong object detection
14
 performance across different IoU thresholds

Mathematical Calculations:

1. Scoring Formula for Vehicle Prioritization

The scoring system dynamically assigns priority scores to detected vehicles based on their
type and count.

The formula is:

S=(3⋅10)+(5⋅5)+(10⋅2)
S =30+25+20
S =75
This score determines the green light priority for the lane.
2. Performance Metrics for Model Evaluation

15
To evaluate the YOLOv8 model used for vehicle detection and classification, the
following metrics are calculated:

APPENDIX
SAMPLE CODE

16
17
OUTPUT SCREENSHOTS

Fig A.3 Realtime Traffic Detection

Fig A.4 Count of instance in dataset

18
Fig A.5 Confusion matrix of the extracted features of the vehicle detection system

19
 REFERENCE

[1] S. G. Rao, "Smart Traffic Management System Utilizing RFID Technology," in Journal of
Urban Traffic Systems, vol. 12, no. 3, pp. 234-241, 2023.

[2] V. Mahavar, "Literature Review on Traffic Control Systems Used Worldwide," in

International Journal of Traffic Optimization, vol. 14, no. 1, pp. 45-52, 2023.

[3] D. Kulkarni, "Intelligent Traffic Surveillance System Using Swarm Technology," in

Proceedings of the International Conference on Smart Cities, pp. 98-105, 2023.

[4] S. Araghi, "A Review on Computational Intelligence Methods for Controlling Traffic Signal
Timing," in Journal of Intelligent Systems Research, vol. 8, no. 4, pp. 198-206, 2023.

[5] H. Wei and G. Zheng, "A Survey on Traffic Signal Control Methods," in Traffic
Engineering and Control, vol. 13, no. 2, pp. 67-74, 2023.

[6] K. Kušić, "A Review of Reinforcement Learning Applications in Adaptive Traffic Signal
Control," in IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 5, pp.
1256-1263, May 2023. DOI: 10.1109/TNNLS.2022.3124567

[7] N. E. Mohamed, "Traffic Light Control Design Approaches: A Systematic Literature

Review," in International Journal of Advanced Traffic Systems, vol. 9, no. 3, pp. 142-150, 2023.

[8] X. R. Lim, "Recent Advances in Traffic Sign Recognition: Approaches and Datasets," in
Computer Vision in Transportation Systems, vol. 19, no. 2, pp. 210-218, 2023.

[9] A. Gupta, "Density-Based Traffic Signal System Using IoT," in Journal of IoT and Smart
City Applications, vol. 7, no. 1, pp. 89-95, 2023.

[10] R. Patel, "Adaptive Traffic Control System Using Deep Learning," in Advances in
Intelligent Transportation Systems, vol. 5, no. 2, pp. 134-141, 2023.

20