Traffic Eye: A Literature Survey on AI-Powered Traffic Rules Violation Detection System

Abstract

The growing number of vehicles on urban roads has turned traffic management and road safety into a pressing challenge. Manual methods of spotting violations are still common, but they tend to be slow, inconsistent, and unable to keep pace with heavy traffic conditions. Recent advances in Artificial Intelligence (AI), Machine Learning (ML), and computer vision provide an opportunity to design smarter, automated solutions for this problem. This work introduces a system that uses the YOLO object detection model to identify traffic rule violations such as speeding, running red lights, and triple riding. To track offenders, Easy OCR is applied for reading vehicle license plates directly from video frames. By bringing together object detection and OCR, the system cuts down on human involvement, lowers the chance of mistakes, and allows for quicker enforcement actions. The approach is intended not only to support authorities in monitoring but also to encourage disciplined driving, making roads safer and traffic management more reliable.

Introduction

The rise in urban vehicle density has increased traffic violations such as speeding, red-light jumping, and improper lane usage, leading to road accidents and congestion. Traditional traffic enforcement relies heavily on manual observation, which is labor-intensive, error-prone, and not scalable.

To address this, Artificial Intelligence (AI), Machine Learning (ML), and smart surveillance technologies are being adopted for automated traffic rule enforcement. These systems utilize real-time video analysis, object detection (e.g., YOLO), and Optical Character Recognition (OCR) to detect and record traffic violations efficiently.

Proposed System Overview

• YOLO (You Only Look Once) is used for high-speed object detection of vehicles and violations.

• Easy OCR/Tesseract OCR enables automatic license plate recognition.

• Integrates live video stream analysis with detection and reporting.

• Aims to reduce manual workload, improve enforcement speed, and enhance road safety.

Literature Survey Highlights

1. Malik et al.

   • Used YOLOv3 with Darknet-53.

   • Detected multiple violations (speeding, red-light jumping).

   • High detection accuracy; needs GPU for scalability.

2. Irani et al.

   • Used CNNs and Tesseract OCR.

   • Detected helmet usage and vehicle classification.

   • Achieved ~75% overall accuracy; struggled with poor image quality.

3. Patil et al.

   • YOLOv5 + Easy OCR.

   • Detected helmet/seatbelt violations; sent email alerts.

   • High accuracy; environmental conditions affected performance.

4. Revadala et al.

   • Built a multi-module Intelligent Traffic Management System (ITMS).

   • Detected phone use, helmet violations, and license plates.

   • Accuracy >93%; challenges with infrastructure and cost.

5. Maduri et al.

   • Used Raspberry Pi and CNNs for helmet/seatbelt detection.

   • Real-time monitoring at highway checkpoints.

   • Achieved >90% accuracy; emphasized low-cost and automation.

6. Charran and Dubey

   • YOLOv4 + Deep SORT for detection/tracking.

   • High precision (98.09%) and plate recognition (99.41%).

   • Offered end-to-end ticketing; validated in real-world tests.

7. Jain et al.

   • YOLOv4 + Easy OCR.

   • Focused on vehicle classification and congestion analysis.

   • Detected traffic density but limited by license plate clarity.

8. Yadav et al.

   • Used unsupervised learning for anomaly detection (K-means, HTM).

   • Detected abnormal driving behaviors (wrong turns, abrupt lane changes).

   • Showed scalability; needs improvement for high-dimensional data.

9. Abishek et al.

   • YOLO + SORT for violation detection and plate recognition.

   • Trained on large, labeled datasets; generated violation reports.

   • Real-time, low-loss detection system.

10. Avupati et al. (2023)

    • YOLOv5 + Haar Cascade.

    • Detected multiple violations; used custom datasets.

    • High mAP (~0.995) for triple riding; struggled with helmet object differentiation.

Key Insights from Literature Survey:

• YOLO-based models (YOLOv3/v4/v5) dominate due to their speed and accuracy.

• License plate recognition is commonly implemented with Easy OCR or Tesseract.

• Helmet, seatbelt, speeding, and signal violations are primary detection targets.

• Deep SORT and SORT are used for real-time tracking.

• Notification systems (email, GPS alerts) and automated ticketing are becoming standard features.

• Challenges include:

  • Environmental variability (e.g., lighting, rain)

  • Poor image quality

  • Need for high-end hardware (GPUs)

  • Difficulty in distinguishing visually similar objects

Conclusion

In today’s experience-driven market, simply offering quality products is no longer enough understanding how customers truly feel has become just as important. This paper explored various research efforts that use artificial intelligence and facial expression recognition to decode human emotions in retail settings. From analysing subtle facial muscle movements to interpreting textual feedback, these systems provide a deeper, more natural understanding of customer satisfaction. Through our study and literature review, it is evident that real-time emotion recognition powered by Convolutional Neural Networks (CNNs), Viola-Jones face detection, and other machine learning methods offers a powerful alternative to traditional feedback systems. These technologies are capable of capturing honest, unbiased emotional responses without requiring verbal input from the customer. They reduce the reliance on manual surveys and help businesses gain faster, data-driven insights into customer preferences and reactions. Ultimately, this paper supports the growing shift toward affective computing—where machines don’t just process data, but understand feelings. By integrating such emotion-aware systems into customer experience strategies, businesses can not only improve satisfaction but also build stronger, more empathetic relationships with their users.

References

[1] Malik, A., Khan, R., Sharma, P., and Mehta, S., “Framework for Automatic Detection of Traffic Violations Using Deep Learning,” Proceedings of International Conference on Intelligent Computing and Computer Vision, 2021. [2] Rani, S., Verma, K., and Gupta, R., “Machine Learning–Based Traffic Violation Detection System for Urban Environments,” International Journal of Emerging Trends in Engineering Research, vol. 11, no. 3, 2023. [3] Patil, V., Deshmukh, A., and Kulkarni, S., “Traffic Rule Violation Detection System Using YOLOv5 and Easy OCR,” International Journal of Scientific Research in Computer Science and Engineering (ICSE), vol. 12, no. 2, 2024. [4] Reveal, R., Naidu, P., and Reddy, M., “Efficient Intelligent-Based Compliance, Detection, Tracking, and Proximity Model for Traffic Systems,” IEEE International Conference on Smart Cities and Systems (ICSC), 2024. [5] Maduri, R., Thomas, J., and Banerjee, K., “Seat Belt and Helmet Detection Using Deep Learning,” International Conference on Artificial Intelligence and Data Engineering (AIDE), 2021. [6] Char ran, S. and Dubey, A., “Two-Wheeler Vehicle Traffic Violations Detection and Automated Ticketing for Indian Road Scenario,” International Journal of Computer Applications, vol. 182, no. 42, 2022. [7] Jain, R., Bansal, P., and Agrawal, N., “Machine Learning-Based Real-Time Traffic Control System,” International Conference on Computing, Power and Communication Technologies (GUION), 2021. [8] Yadav, V., Singh, M., and Kumar, A., “Detection of Anomalies in Traffic Scene Surveillance Using Unsupervised Learning,” IEEE International Conference on Intelligent Transportation Systems, 2018. [9] Abishek, R., Nair, S., and Prasad, T., “Detection of Traffic Violation and Vehicle Number Plate Using Computer Vision,” International Journal of Scientific & Technology Research (INSTR), vol. 13, no. 1, 2024. [10] Aquatic, M., Raghavan, H., and Srinivas, K., “Traffic Rules Violation Detection Using YOLO and Haar Cascade,” International Conference on Artificial Intelligence and Smart Systems (IC AIS), 2023.

Copyright

Copyright © 2025 Bhavana B M, Mr. Sathyanarayana S, Bhavana M K, G N Nisarga, Harshitha S. 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.