AI-Powered Vehicle Speed Detection and Violation Monitoring System

Abstract

Road safety remains one of the most important challenges in modern transportation systems. Over-speed driving contributes significantly to road accidents, fatalities, traffic congestion, and property damage. Conventional traffic speed monitoring methods such as radar guns, loop detectors, and manual supervision are expensive, infrastructure dependent, and difficult to deploy at scale. This paper presents an intelligent surveillance framework for automated vehicle speed detection and violation monitoring using deep learning and computer vision. The proposed system is based on the real project implementation developed using Python, OpenCV, PyTorch, and Ultralytics YOLOv8 Nano. Vehicles such as cars, motorcycles, buses, and trucks are detected in real time from CCTV or uploaded traffic footage. Multi-object identity continuity is maintained through the integrated tracking pipeline. Vehicle speed is estimated by analyzing centroid displacement over time using a calibrated pixel-to-meter conversion ratio. A temporal smoothing strategy is applied to improve speed stability and reduce jitter caused by frame-level noise. The implemented system uses a threshold of 60 km/h to automatically identify over-speed vehicles. Violating vehicles can be highlighted, logged, and integrated into a backend dashboard for monitoring authorities. The framework offers a practical lowcost alternative to dedicated radar infrastructure and demonstrates the ability to convert ordinary surveillance cameras into intelligent traffic sensors. The solution is suitable for smart city deployments, road safety analytics, and scalable transportation governance.

Introduction

The text describes a real-time AI-based vehicle speed detection system designed to improve road safety in rapidly growing urban transportation networks.

It highlights that traditional speed monitoring methods like radar guns and road sensors are expensive, limited in coverage, and require manual effort. To overcome these issues, the proposed system uses existing CCTV infrastructure combined with Artificial Intelligence and Computer Vision to automatically detect vehicles, estimate their speed, and identify overspeed violations.

The system is built using YOLOv8 Nano for vehicle detection, multi-object tracking, and OpenCV-based video processing. It performs real-time functions such as vehicle detection, speed estimation, violation detection (using a 60 km/h threshold), and backend logging for smart city integration. Speed is calculated using pixel-to-distance calibration to convert motion in video frames into real-world speed.

Previous research in traffic monitoring is reviewed, showing a shift from traditional image processing methods (like background subtraction and optical flow) to deep learning approaches such as CNNs and YOLO. The text also discusses improvements in tracking methods like DeepSORT and ByteTrack.

The system is tested on diverse traffic videos (urban roads, highways, intersections) under different conditions. It uses preprocessing steps like frame extraction, resizing, and noise handling to improve accuracy. Challenges such as occlusion, motion blur, weather effects, and lighting changes are also noted.

Conclusion

This paper presented an accurate documentation of the real implemented project titled Intelligent Surveillance: AIPowered Vehicle Speed Detection and Violation Monitoring System. The framework combines YOLOv8 Nano detection, tracking-based motion continuity, centroid displacement speed estimation, temporal smoothing, and threshold-based violation monitoring. The real uploaded source code demonstrates that effective traffic analytics can be achieved using software intelligence layered on top of existing camera infrastructure. By reducing dependence on expensive hardware sensors, the system offers a scalable and affordable approach for future road safety enforcement and smart transportation systems. This paper presented an effective and practical intelligent surveillance framework for AI-powered vehicle speed detection and violation monitoring using computer vision and deep learning techniques. The developed system successfully integrates YOLOv8 Nano for real-time vehicle detection, tracking-based motion continuity, centroid displacement speed estimation, temporal smoothing, and configurable thresholdbased violation recognition. The implementation demonstrates that existing CCTV infrastructure can be transformed into smart traffic sensors without requiring expensive dedicated radar or loop detector hardware. By using a software-driven approach, the proposed framework offers a low-cost, scalable, and flexible solution for modern transportation management. Experimental observations confirmed reliable multi-vehicle detection, stable speed estimation, and practical real-time monitoring capability under calibrated camera environments. The use of a 10-frame measurement window and weighted smoothing significantly improved consistency of displayed speed values while reducing noise caused by frame-level variations. Another major contribution of this project is its extensible architecture. The developed model can be integrated with backend dashboards, historical analytics systems, and automated evidence logging platforms for traffic authorities. This creates opportunities for centralized road safety governance and data-driven decision making. Overall, the proposed system demonstrates how artificial intelligence can be effectively applied to public safety applications. With future enhancements such as automatic number plate recognition, lane discipline analytics, night-time optimization, and cloud deployment, the framework can evolve into a complete next-generation smart traffic surveillance platform.

References

[1] Ultralytics, “YOLOv8 Documentation,” 2025. [2] J. Redmon et al., “You Only Look Once: Unified Real-Time Object Detection,” CVPR, 2016. [3] OpenCV Foundation, “OpenCV Library Documentation,” 2025. [4] PyTorch Team, “PyTorch Framework,” 2025. [5] N. Wojke et al., “Simple Online and Realtime Tracking,” IEEE ICIP, 2017. [6] Y. Zhang et al., “ByteTrack: Multi-Object Tracking by Associating Every Detection Box,” ECCV, 2022. [7] IEEE ITS Society, “Smart Transportation Systems Overview,” 2024. [8] Deep Learning Applications in Traffic Surveillance, IEEE Surveys, 2024. [9] Computer Vision for Smart Cities, International Journal Review, 2023. [9] Automated Speed Estimation from CCTV Video Streams, Research Trends, 2024. [10] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards RealTime Object Detection with Region Proposal Networks,” NeurIPS, 2015. [12] W. Liu et al., “SSD: Single Shot MultiBox Detector,” ECCV, 2016. [11] A. Bewley, Z. Ge, L. Ott, F. Ramos, and B. Upcroft, “Simple Online and Realtime Tracking,” ICIP, 2016. [12] A. Bochkovskiy, C. Wang, and H. Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” arXiv preprint arXiv:2004.10934, 2020. [15] G. Jocher et al., “YOLO by Ultralytics,” GitHub Repository, 2024. [13] D. Comaniciu and P. Meer, “Mean Shift: A Robust Approach Toward Feature Space Analysis,” IEEE TPAMI, 2002. [14] R. Kalman, “A New Approach to Linear Filtering and Prediction Problems,” Journal of Basic Engineering, 1960. [15] M. Everingham et al., “The Pascal Visual Object Classes Challenge,” IJCV, 2010. [16] T. Lin et al., “Microsoft COCO: Common Objects in Context,” ECCV, 2014. [17] Intelligent Transportation Analytics Report, “AI-Based Road Safety Monitoring Systems,” Technical Review, 2025.

Copyright

Copyright © 2026 Padala Vamsi, Vayala Pavani, Dr. Konda Chaitanya, Palaparthi Swapna. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.