- Download 14
- File Size 374.21 KB
- File Count 1
- Create Date 11/06/2025
- Last Updated 11/06/2025
Design and Implementation of a Real-Time Vehicle Overspeed Detection System Using YOLOv8, Optical Character Recognition, and Dual-Camera Analytics
Devanshu Vijay Patil, Nehal Mahendra Rane
Devanshu Vijay Patil, Computer Engineering, Zeal College of Engineering and Research
Nehal Mahendra Rane, Computer Engineering, Zeal College of Engineering and Research
Abstract - The review of methodologies regarding real-time vehicle speed detection, estimation, and tracking using some advanced computer vision techniques, using modern deep learning techniques, by applying models like YOLOv8, Deep SORT, and GMM by possible application for the accurate detection and tracking of vehicles. Thus, the research refers to fewer techniques that have achieved a Mean Absolute Error of 35 and RMSE of 422 relating to the estimation of speed; it refers to the addition of better time interpolation methods, including vehicle acceleration that lowers the error in the speed estimation to 07% levels. Another discussion incorporated in the paper is the high-resolution dataset for fine-grained vehicle recognition and Classification, Low-Cost Speed Detection Systems that rely on frame difference methods and IoT integration These innovations enhance the real-time surveillance ability especially in sensitive areas such as schools and hospitals, as it is efficient, although some challenges brought about by environmental factors such as lighting conditions, angles, and other cameras bring about difficulties. Below are the results from the survey and indicate in what ways this could be attributed to developing better traffic management, police service, and intelligent transportation systems. Results Specifically, these reveal the feasibility of scalable, accurate, and cost-effective solutions towards road safety and effective traffic management. The fast pace of urbanization, coupled with more and more cars on the road, has resulted in an enormous surge in traffic violations, especially over speeding, that is one of the leading causes of road accidents. Current speed checking devices, like radar guns and single speed sensors, can only detect car speed at a point. This method is prone to cheating by drivers who reduce speed for a short while at familiar locations, thus evading detection. Thus, there is an urgent need for a more secure and safer system to assess car speed over a long distance.
This project presents a real-time overspeed detection system with a two-camera architecture to estimate the average speed of a vehicle between two static observation points. Based on state-of-the-art CV & DL processes, the system is precise and reliable in detection. Vehicle detection is performed using the YOLOv8 (You Only Look Once, Version 8) object detection model, which is precise and efficient. License plate recognition is done by Tesseract OCR, which reads alphanumeric characters from identified vehicle plates to uniquely recognize a vehicle.
The timestamps are captured by the system when the vehicles cross the entry and exit points. The average velocity is obtained using the known distance between the two cameras. If the resultant speed is above the pre-configured threshold (for instance, 100 km/h), the vehicle is detected automatically for the speed limit violation. Detailed information, such as license plate number, speed, and detection time, is reflected on an interactive real-time dashboard, which is developed using Streamlit.
The proposed solution is economically efficient, hardware resource-saving, and suitable for installation in highway and urban environments. Its modularity also renders it easy to integrate into intelligent traffic monitoring systems and smart city infrastructures, thus maximizing the enforcement of traffic regulations and encouraging the enhancement of road safety in general.
Key Words: Vehicle Overspeed Detection, YOLOv8, Tesseract OCR, Average Speed Monitoring, License Plate Recognition, Dual-Camera System, Real-Time Traffic Surveillance, Streamlit Dashboard, Deep Learning, Smart City Infrastructure