Traffic Violation Detection System: Overspeed, Signal Jump, No Helmet, Triple Ride

Abstract

In recent time surveys, the deaths and injuries due to traffic violations have increased chiefly on Indian roads. This necessitates the assistance of an automated computer vision-based object detection model, as manually identifying vehicles violating traffic rules is inefficient. This paper presents a Traffic Violation Detection System capable of detecting multiple violations — including overspeed, red-light signal jumping, no helmet, and triple riding — using single video frames from surveillance cameras. The input video stream is processed and annotated to carry out multiple detection processes simultaneously. The dataset used for red-light jumping detection is COCO, while the dataset for overboarding is created by annotating images obtained from Google. The YOLOv7 deep learning model is trained and its output is visualized using TensorBoard. Performance parameters include Precision, Recall, F-measure, and P-measure. The system achieves 93% accuracy for red-light signal violation detection, and a mean average precision (mAP) value of 0.5:0.95 for overboarding detection. The system is implemented using Python, PyTorch, and OpenCV, and is containerised for ease of deployment. A user evaluation confirms reliable detection performance, demonstrating measurable improvement over traditional manual traffic monitoring.

Introduction

The text describes an AI-based traffic violation detection system designed to improve road safety by automating the identification of violations from surveillance video feeds. It addresses the limitations of manual monitoring, which is slow, inconsistent, and unable to handle multiple violations simultaneously.

The proposed system uses deep learning, specifically the YOLOv7 object detection model, to detect multiple violations such as red-light jumping, helmet non-compliance, triple riding, and overspeeding in real time. It is trained using a combination of the COCO dataset and custom-annotated datasets to improve accuracy and adaptability. Performance is evaluated using standard metrics, achieving about 93% accuracy for red-light violation detection.

Existing systems are limited because they typically focus on a single violation type, struggle with real-time processing under varying environmental conditions, and lack integrated frameworks. The proposed solution overcomes these issues by enabling unified, real-time, multi-violation detection within a single system.

Key contributions include:

• A unified framework for detecting multiple traffic violations simultaneously
• Use of YOLOv7 for real-time object detection
• Integration of standard and custom datasets for better generalization
• Improved automation to reduce dependence on manual traffic enforcement
• Enhanced road safety through faster and more reliable violation detection

Overall, the system aims to provide a scalable, efficient, and real-time solution for automated traffic monitoring and enforcement.

Conclusion

This paper presented a Traffic Violation Detection System that demonstrates the effective application of computer vision and deep learning techniques in automating traffic monitoring. The system accurately detects multiple violations — red-light jumping, overspeeding, triple riding, and riding without helmets — using real-time video streams from surveillance cameras. By integrating the YOLOv7 object detection model with OpenCV-based processing, the system achieves high accuracy and efficient performance. The development process involved dataset collection, preprocessing, annotation, model training, and evaluation. The use of both the COCO dataset and custom-annotated datasets improved the system\'s ability to generalize across different traffic scenarios. Performance evaluation using accuracy, precision, recall, F1-score, and mAP confirmed that the model performs reliably. One of the key achievements is simultaneous multi-violation detection from a single video stream, significantly improving monitoring efficiency over traditional methods.

References

[1] J. Chiverton, \"Helmet presence classification with motorcycle detection and tracking,\" Intelligent Transport Systems (IET), vol. 6, no. 3, pp. 259-269, September 2012. [2] Gomathi et al., \"Automatic Detection of Motorcycle without Helmet using IoT,\" South Asian Journal of Engineering and Technology, 2017. [3] S. Ibadov et al., \"Algorithm for detecting violations of traffic rules based on computer vision approaches,\" MATEC Web of Conferences, vol. 132, 05005, 2017. [4] A. C. P. Uy et al., \"Automated Traffic Violation Apprehension System Using Genetic Algorithm and Artificial Neural Network,\" IEEE Region 10 Conference, 2016. [5] A. Narkhede et al., \"Automatic Traffic Rule Violation Detection and Number Plate Recognition,\" IJSTE, vol. 3, issue 09, March 2017. [6] K. Dahiya, D. Singh, and C. K. Mohan, \"Automatic Detection of Bike-riders without Helmet using Surveillance Videos in Real-time,\" IJCNN, 2016. [7] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, \"You Only Look Once: Unified, Real-Time Object Detection,\" CVPR, 2016. [8] K. He, G. Gkioxari, P. Dollar, and R. Girshick, \"Mask R-CNN,\" Facebook AI Research (FAIR), January 2018. [9] C.-Y. Wang, A. Bochkovskiy, and H.-Y. M. Liao, \"YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors,\" CVPR, 2023. [10] P. Bhaskar et al., \"Image Processing Based Vehicle Detection and Tracking Method,\" Universiti Teknologi PETRONAS, Malaysia. [11] A. Riza et al., \"Localization of License Plates Using Optimized Edge and Contour Detection Technique,\" IEEE TENCON, 2017. [12] P. S. Ha and M. Shakeri, \"License Plate Automatic Recognition based on Edge Detection,\" International Journal of Computer Science, 2012. [13] A. Asim et al., \"A Vehicle Detection Approach using Deep Learning Methodologies,\" IEEE Access, 2021.

Copyright

Copyright © 2026 Mr. Nakka Satyananda Varma, Mr. Kuruva Purna chand, Mr. Thurumalla Sai Rajendra, Mr. Muramalla Janardhan. 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.