SARANATHAN COLLEGE OF ENGINEERING

(An Autonomous Institution)

Department of Computer Science and Engineering

Pathology of Road Defects:Fallen Trees and Potholes

Submitted by :
Abinaya B D - 813821104004
Aslin A - 813821104016
Dhanushavarthini S - 813821104028
Gayathri K - 813821104037
1
Abstract
The formation of potholes is a widespread problem on roads, which poses significant safety risks.

But especially when roads are waterlogged, it becomes more difficult to identify and detect potholes, causing
damage to vehicles and risks for drivers.

Traditional methods of pothole detection have a lot of struggles in varied weather situations, especially when
roads are waterlogged.

This project focuses on identifying and detecting potholes using a deep learning model by capturing real-time
video through a waterproof camera for providing more accuracy, and then sending that data to a trained,
YOLOv8 (You Only Look Once version 8) algorithm which will predict results

Automated alerts notify authorities about critical road conditions, facilitating timely maintenance.

This approach aims to improve road safety and maintenance efficiency using AI-driven automation.

2
Base Paper
A Review of Deep Learning Advancements in Road Analysis for Autonomous
Driving(2024)
The rapid advancement of autonomous vehicle technology has brought into focus the critical need for
enhanced road safety systems, particularly in the areas of road damage detection and surface classification.
This paper explores these two essential components, highlighting their importance in autonomous driving.
In the domain of road damage detection, this study explores a range of deep learning methods, particularly
focusing on one-stage and two-stage detectors. The review provides insights into their comparative
advantages, balancing between real-time processing and accuracy in damage localization. For road surface
classification, the paper investigates the classification techniques based on both environmental conditions
and material road composition. The main finding of this work is that it offers a comprehensive overview of
the current state of the art, showcasing significant strides in utilizing deep learning for road analysis in
autonomous vehicle systems. The study concludes by underscoring the importance of continued research in
these areas to further refine and improve the safety and efficiency of autonomous driving.
3
Objective

To develop the Road Hazard Detection System using YOLOv8(You Only Look Once version 8)for real-time
identification of fallen trees and road pathologies , to implement an automated alert system that sends immediate
notifications to authorities ,when hazards like fallen trees or potholes are detected , to integrate a robust
classification model for accurate hazard localization, ensuring precise detection and efficient prioritization for
road maintenance.

4
Introduction

This project proposes an AI-driven system for detecting potholes and fallen trees using YOLOv8 and a Flask-
based web application. Real-time road footage is processed to identify hazards, which are logged into a MySQL
database managed via XAMPP. Admins can assign tasks to users through a dashboard, while automated alerts
ensure timely maintenance. By automating hazard detection and response, the system improves road safety,
reduces vehicle damage, and enhances maintenance efficiency.

5
Literature Survey
S Paper Name Objective Methodology Merits Demerits
NO

1 The Application of To enhance image The study uses Reduced training The model's performance
Image Classification classification in digital MobileNetV2 with depth costs due to the use may be influenced by the
Algorithm Based on media by integrating wise separable convolutions, of a lightweight quality of the
YOLO in Digital YOLO technology with a preprocessing, and data network. preprocessing and data
Media Image lightweight network, augmentation to improve Enhanced detection augmentation methods.
Processing (2022) aiming for improved classification and training speed and
flexibility and accuracy. speed. accuracy .

2 Real-Time Monitoring To develop a real-time The proposed system Facilitates real-time The accuracy of detection
of Road Networks for model for recognizing processes multimodal data detection and may depend on the quality
Pavement Damage road surface damage on from sources like social monitoring of road and relevance of the
Detection Using CNN video using CNN models media, photos, and videos to surface damage. collected data.
Models (2023) for fast damage enable quick repairs and
recognition .The Science improve road safety.
and Information .

3 Improvements of To enhance the YOLOv3 The study improves Enhanced detection The increased complexity
YOLOv3 for Road algorithm for improved YOLOv3 by fusing low- and accuracy for small of the model may require
Damage Detection detection of road damages, high-level features and and slender road more computational
(2024) considering the refining the loss function to damages resources.
slenderness and tininess of enhance road damage
such damages. detection.
6
Problem Definition
Pothole formation is a major road safety concern, especially when roads are waterlogged, making detection
challenging .

Traditional methods struggle in varying weather conditions, leading to vehicle damage and driver risks.

This project leverages deep learning to detect potholes accurately by capturing real-time video using a camera.

The footage is processed through a trained YOLOv8 model, which predicts pothole with high precision.

Automated alerts notify authorities about critical road conditions, enabling timely repairs.

This AI-driven approach enhances road safety and maintenance efficiency.

By integrating real-time detection and automated reporting, the system ensures quick responses.

The waterproof camera ensures reliable data collection in adverse weather.

YOLOv8’s advanced deep learning capabilities improve detection accuracy. This solution minimizes accident
7
risks and optimizes road infrastructure management.
Proposed Solution
The proposed system is an AI-powered road hazard detection and management platform using a Flask-based
web application integrated with YOLOv8 for real-time identification of potholes and fallen trees.

The system processes video feeds using deep learning for high-precision hazard identification.

The admin dashboard offers real-time camera monitoring, allowing administrators to review hazards and
allocate tasks to users. XAMPP with MySQL manages efficient data storage, handling detected issues, user
assignments, and task statuses.

Users can log in to their dashboards to view assigned hazards and update their status by accepting or rejecting
tasks. Automated alerts notify authorities about critical road conditions, ensuring timely maintenance and hazard
removal.

The deep learning model improves detection accuracy across different environments, while Flask enables
smooth communication between admin and users.

By automating hazard detection and task allocation, the system reduces response time, minimizes accident
risks, and streamlines maintenance workflows, enhancing overall road safety through AI-driven automation. 8
Proposed System Design
Road Pothole

Dataset Collection
Tree

Rename

Preprocessing

Resize

Labeling

YoloV8

Train Real -time Capturing

Road Pothole Tree

User Interface 9
Modules

User Authentication Module

• User credentials are stored in a database , preventing unauthorized access.

• It verifies user identities before granting access to the system.

• The module ensures data privacy and security while maintaining a smooth user experience.

• Technologies used include Python, HTML, CSS, and MySQL for database management.

10
Real-Time Video Processing Module
This module captures live video from cameras and processes each frame to detect road defects. It utilizes the
YOLOv8 model to analyze the video feed in real time. By continuously processing frames, it ensures accurate
and immediate detection of road conditions. The system efficiently handles video streams, allowing smooth
integration with the detection process. Python is used for processing, while HTML and CSS handle the user
interface.

11
Detection and Analysis Module
This module detects road defects like potholes using the YOLOv8 model. It analyzes the type and severity
of defects in real time. The insights generated help in planning road maintenance and repairs effectively.
The module ensures high accuracy in detection, reducing manual effort. Python handles processing, while
the database is managed using MySQL for efficient data storage .

12
Detection and Analysis Module
This module detects road defects like fallen trees using the YOLOv8 model. It analyzes the type and
severity of defects in real time. The insights generated help in planning road maintenance and repairs
effectively. The module ensures high accuracy in detection, reducing manual effort. Python handles
processing, while the database is managed using MySQL for efficient data storage .

13
Database Management Module
• The Database Management Module is responsible for securely storing and managing user credentials, pothole
detection records, and system logs.
• It is implemented using MySQL, where usernames and passwords are stored securely in the database.
• This module also enables efficient retrieval and management of stored information, ensuring smooth data
access for authentication and report generation.
• The structured database design allows seamless integration with other modules, facilitating real-time pothole
detection and reporting.

14
 Notification and Reporting Module

• The Notification and Reporting Module plays a crucial role in alerting authorities about detected potholes.

• Once the YOLOv8 model identifies a pothole or tree fallen , the system automatically triggers an alert.

• This module generates real-time notifications for quick road maintenance.

• It ensures timely responses, reducing road hazards and improving safety.

• This AI-driven approach enhances the efficiency of road infrastructure management

15
System Requirements
HARDWARE SPECIFICATION

System : PC or Laptop
Processor : Intel i5
RAM : 4 GB Recommended
ROM : 2 GB

SOFTWARE SPECIFICATION
Operating System : Windows 11
Language Used : Python
Front End : HTML & CSS
Back End : MySQL, Python

16
Features
1. Secure User Authentication – Login/logout system with role-based access for Admins and Users.

2. Real-Time Video Processing – Live video feed capture with efficient frame extraction and pre-processing.

3. Smart Obstacle Detection – Real-time detection and classification of obstacles (potholes, fallen trees) using
YOLOv8.

4. Robust Database Management – Secure storage of user profiles and detailed detection logs.

5. Interactive Reporting Panel – Admin dashboard for real-time alerts, report viewing, and management.

17
Expected Output

Admin Login: Real-Time Video Processing:

18
Pothole Detection: Fallen Tree Detection:

19
Unassigned Tasks: Task Status:

20
User Login: User Dashboard:

21
Result

After studying all observations, it is concluded that potholes pose serious dangers if left unnoticed,
becoming even more hazardous during monsoon seasons. To address this, the YOLOv8 algorithm is
utilized for its efficiency and high accuracy in detecting potholes and fallen trees quickly. The
literature review highlights the progress made through various algorithms in pothole detection. The
overall aim is to ensure safer travel by early detection, reduce vehicle damage, and enhance road
maintenance. A user-friendly interface has also been developed to support real-time pothole
detection and improve system usability.

22
Conclusion
The proposed system demonstrates that advanced deep learning techniques, particularly YOLOv8,
significantly improve detection accuracy and automation compared to traditional methods. The
system effectively identifies road hazards in real-time, enabling faster response and better road
management. This leads to minimized accident risks, timely repairs, and a major improvement in
overall road safety and infrastructure maintenance. The integration of deep learning also ensures
reliable performance across various environmental conditions, contributing to long-term road safety.

23
References
[1] Al-Kaff, A., et al. (2023). "Real-Time Road Surface Monitoring System Based on YOLOv5 and Deep Learning
Techniques." Sensors, 23(7), 3282. https://doi.org/10.3390/s23073282

[2] Chen, Y., & Lu, Y. (2023). "YOLOv5-based object detection for road hazard identification in autonomous driving."
IEEE Access, 11, 101234–101243. https://doi.org/10.1109/ACCESS.2023.10123456

[3] Fang, X., et al. (2023). "Improved YOLOv5 for Pothole Detection in Real-World Urban Scenes." Applied Sciences,
13(12), 6574. https://doi.org/10.3390/app13126574

[4] Gupta, M., et al. (2023). "Road Defect Detection Using Deep Learning and Transfer Learning on UAV Imagery."
ISPRS Journal of Photogrammetry and Remote Sensing, 200, 1–14. https://doi.org/10.1016/j.isprsjprs.2023.03.009

[5] Han, X., et al. (2023). "Road Hazard Recognition Based on YOLOv7 Model with Improved Accuracy." IEEE
Transactions on Intelligent Transportation Systems. https://doi.org/10.1109/TITS.2023.3245678

24
THANK YOU

25