Volume 13, Issue 5, May 2024

Impact Factor: 8.423

 | DOI:10.15680/IJIRSET.2024.1305486 |

Real Time Object Detection for Traffic

Analysis using YOLOv8
Ms Dimple R1, Hemanth Gowda G N2, Sindhu D M3, Supreetha G J4, Anish MU5
Assistant Professor, Department of Electronics and Communication Engineering, PES College of Engineering,
Mandya, India1
Final Year UG Student, Department of Electronics and Communication Engineering, PES College of Engineering,
Mandya, India2-5

ABSTRACT: In modern transportation systems, real-time monitoring and analysis of traffic scenarios are important
for efficient traffic management and safety enhancement. This paper give out a novel approach to real-time object
detection tailored for traffic analysis, with a focus on speed detection and lane change detection. Our system utilizes
state-of-the-art deep structural learning techniques to accurately detect and track vehicles in live video streams from
traffic cameras. Through the integration of advanced algorithms, our system is capable of not only identifying vehicles
but also determining their speeds and detecting lane changes in real time. Experimental results demonstrate the
effectiveness and efficiency of this approach in various traffic scenarios, showcasing its potential for deployment in
intelligent transportation systems for enhanced traffic management and safety enforcement.

KEYWORDS: YOLOv8, Speed Detection, Object Detection, Vehicle Counting, Computer vision.

I. INTRODUCTION

Modern urban environments are experiencing unprecedented growth in population and vehicular traffic. As our cities
expand, so does the complexity of managing and optimizing traffic flow. It’s really important to understand and study
traffic patterns effectively and quickly. Accurate and real-time information about traffic patterns can lead to smarter
urban planning, improved safety, and enhanced overall quality of life. Traditional methods of traffic analysis often fall
short in providing the speed, precision, and scalability required for today's dynamic traffic scenarios. This is at which
the integration of cutting-edge technologies becomes indispensable.

Our project mainly focused on utilizing the YOLO (You Only Look Once) for object detection to revolutionise the way
we analyse traffic. The selection of this problem stems from a convergence of technological advancements and societal
needs. YOLO, a state-of-threat real-time object detection algorithm, has demonstrated the exceptional work in different
fields, that includes computer vision and image processing. Its capability to analyze images in real-time with great
accuracy makes it a standout choice for numerous applications, including traffic analysis. The decision to apply YOLO
for traffic analysis is rooted in the recognition that traditional methods often face challenges in handling diverse and
dynamic traffic scenarios. YOLO's unique approach of dividing an image to a grid and simultaneously forecasting the
bounding boxes and class probabilities for each divided cell offers a nimble and efficient solution. Our aim is to harness
the power of YOLO to precisely detect and track vehicles, pedestrians, and other relevant objects in a traffic scene. By
doing so, we aim to give valuable perception into traffic patterns, congestion, and potential safety hazards, empowering
urban planners and traffic management authorities to make informed decisions.

II. LITERATURE REVIEW

 The study referred from paper[1] authors proposed a model using “YOLO and SORT for the estimation of
traffic flow”. The purpose of this paper is traffic flow approximation using low-quality CCTV video footage in
real-time. They primarily concentrated on multi-lane road. They trained the novel YOLOv4 algorithm for five
object classes such as car, truck, van, bike and bus . They developed an algorithm using a sort tracker to count
vehicles based on their movement direction like Northbound and Southbound
 In this paper[2] the authors proposed a model using “Computer vision and Deep Learning for the lane and
vehicle detection”. This paper gives a lane-vehicle detection and tracking system. The authors divided this project
into two steps they are vehicle detection and Lane detection. For vehicle detection they used YOLO Framework.
For the lane detection they use many Image manipulation methods using color spaces.

IJIRSET©2024 | An ISO 9001:2008 Certified Journal | 9487

 | DOI:10.15680/IJIRSET.2024.1305486 |

 In this paper[3] authors introduced a “Efficient Roundabout Supervision: Real-Time Vehicle Detection and
Tracking on Nvidia Jetson Nano”. They installing a smart camera on the primary route connecting the two cities.
They implemented the system using the YOLOv7 tiny object detection model to detect and classify the different
kinds of vehicles crossing the main road.
 In this paper[4] authors proposed a “Object detection and tracking using deep learning algorithm” Detection
and identification of smaller objects. To generate a boundary boxes around the vehicles sliding window and roll
polling methods are used. COCO dataset is used for object classification. It shows the name of the each and every
object. Consumes more time to detect and classify the object
 In this paper[5] author proposed a “Image manipulating methods and challenges”. Noise reduction, image
enhancement, object recognition, and segmentation. Techniques: Image pre-processing, Rotation, Image
restoration, Image analysis.

III. PROPOSED WORK AND OBJECTIVES

To develop an effective object detection system for a traffic analysis using YOLO algorithm.
The block diagram in Fig.1 The system works by first capturing the video of the road. This video image is then
supplied into a detector, which was referred to identify lane markings in the picture. The detector may use a variety of
techniques to do this, such as edge detection or image segmentation. After the lane markings identifications had done,
the system can then track the movement of the lanes over time. This information can be used to count the vehicles on
the road, or to detect rate of change of movement of the vehicles.

Fig 1: Block Diagram of object detection using YOLOv8 Algorithm

IJIRSET©2024 | An ISO 9001:2008 Certified Journal | 9488

 | DOI:10.15680/IJIRSET.2024.1305486 |

Video source: The video footage provides the raw data for traffic analysis, capturing real-time information about
vehicles and their movements on the road. Image segmentation: By segmenting the image, the system can isolate
vehicles from the background, allowing for accurate vehicle detection and tracking.

Edge detection: Detecting edges helps identify lane markings and boundaries, which are difficult for understanding lane
configurations and vehicle positioning within lanes.

Hough transform: Detecting lane lines using the Hough transform enables the system to determine the layout of the
road and monitor lane occupancy, which is essential for analyzing traffic flow and congestion .Object detection
(YOLOv8): Detecting vehicles using YOLOv8 provides data on vehicle presence, spatial distributions, count and
offering insights into traffic density and congestion levels.

Lane Tracking :Tracking lane lines over time helps monitor lane changes, merge behaviours, and lane occupancy
rates, providing valuable information for traffic flow optimization and congestion management. Perspective
transformation: Adjusting for perspective distortion ensures accurate measurements and analysis of vehicle speeds,
distances, and trajectories, essential for understanding traffic dynamics and behaviour. Background subtraction: By
isolating moving vehicles from the background, background subtraction enables the system to give importance to
dynamic vehicles of the scene, such as vehicles in motion, facilitating accurate traffic flow analysis.

Detected line and detected vehicle: Outputs from the system, such as detected lane lines and vehicles, serve as the basis
for traffic analysis, providing data on lane occupancy, vehicle distribution, and traffic patterns.

Count vehicle module: Counting of vehicles on road facilitates traffic volume estimation, helping quantify traffic flow
rates and congestion levels at specific locations and times.

Estimate vehicle speed: Enables the system to analyse traffic dynamics, identify areas of congestion, and assess the
effectiveness of traffic management strategies, such as speed limits and lane configurations

IV. RESULTS

The YOLO Algorithm for Traffic Analysis provides comprehensive results like vehicle counting, speed detection,
bounding box coordinates and class labels. It accurately detects and counts vehicle in give frame.

Fig 2: Before Segmentation

IJIRSET©2024 | An ISO 9001:2008 Certified Journal | 9489

 | DOI:10.15680/IJIRSET.2024.1305486 |

Fig 3: Vehicle count and Speed detection

Fig 4: Lane change detection of vehicles on road

V. CONCLUSION

YOLOv8 is a deep learning algorithm with object detection and tracking capabilities. It makes counting and tracking
automobiles simple. YOLOv8's strong ability to detect and track objects makes it a handy tool for keeping an eye on
traffic. Advantages
1. It can detect the small objects.
2. By monitoring the density of the vehicles in different areas, traffic authorities can better manage and optimize
traffic flow.
3. Object detection is not limited to vehicles, it can also identify pedestrians.

IJIRSET©2024 | An ISO 9001:2008 Certified Journal | 9490

 | DOI:10.15680/IJIRSET.2024.1305486 |

REFERENCES

[1]. Nilani Algiriyage, Raj Prasanna, Emma E H Doyle, Kritin Stock proposed a “Towards Real-Time Traffic Flow
Estimation using YOLO and SORT from Surveillance video Footage”, International journal of engineering and
advanced technology (IJEAT), volume8, issue-3, Feb 2019.
[2]. Sunil Kumar, Bipin karki ,Maryam Mahmood, Soumya Visnoi,gauri Shivhare, proposed a “Lane and Vehicle
detection using computer vision and Deep Learning”, International journal of recent technology and engineering,
(IJRTE) ISSN: 2277- 3878, volume-8, issue-1, May 2019.
[3]. Imane Elmanaa, My Abdelouahed Sabri, Yassine Abouch, proposed a “Efficient Roundabout.
Supervision: Real-Time Vehicle Detection and Tracking on Nvidia Jetson Nano”, International journal of scientific
research and engineering, volume-5, issue-2, 2019.
[4]. Cheng-Jian Lin,Shiou-YunJeng and Heng-weiLioa, proposed a “A Real-Time Vehicle Counting, Speed Estimation
and Classification System Based on virtual Detection Zone and YOLO”, International journal of computational
intelligence and informatics, volume-5, NO-1, 2018.
[5]. Deepika B, Dorthy Shaji, proposed a “Object detection and tracking using deep learning algorithm”, Intelligent
systems research, volume163, 2018.
[6]. [Link], Bagyamani, proposed a “Traffic analysis on highways based on image processing” Journal of
emerging technologies and innovative research, volume-9, issue-9, September 2022.
[7]. Akshay Chaturi Goswami, proposed a “Image processing techniques and challenges”, Journal of emerging
technologies and innovative research, volume9, issue-9, September 2022.
[8]. Hutesh Kumar Gautam and Ramesh Kumar Mohapatra, proposed a “Speed prediction of fast approaching vehicle
using moving camera”, International research journal of engineering and technology, volume-9, issue-5, may 2022.

IJIRSET©2024 | An ISO 9001:2008 Certified Journal | 9491