Deep Learning-Based Real-Time Detection of Helmet Non-Compliance and Triple Riding Violations in Traffic Surveillance Systems

Abstract

The rapid growth of urbanization and motor vehicle usage has led to an increase in traffic violations, particularly in developing countries where enforcement mechanisms are often limited. Among these violations, helmet non-compliance and triple riding on twowheelers are significant contributors to road accidents and fatalities. Traditional traffic monitoring systems rely heavily on manual supervision, which is prone to inefficiencies, human error, and scalability challenges. To address these limitations, this research proposes a deep learning-based automated traffic violation detection system capable of identifying helmet violations and triple riding in real time.The proposed system leverages the YOLOv8 (You Only Look Once version 8) object detection framework to detect persons, motorcycles, and helmets from video streams. The system processes frames from video input and identifies relevant objects based on pre-trained models. A novel approach using Intersection over Union (IoU) is employed to associate detected persons with helmets and motorcycles, enabling accurate identification of violations. If a person is detected without a corresponding helmet, the system flags it as a helmet violation. Similarly, when more than two persons are associated with a single motorcycle, the system identifies it as a triple riding violation. To ensure traceability and enforcement, the system captures and stores visual evidence of each violation along with timestamps in a structured log file. This data can be further utilized for legal or administrative purposes. The implementation is carried out using Python, OpenCV, and the Ultralytics YOLO framework, ensuring high computational efficiency and real-time processing capability.Experimental results demonstrate that the proposed system achieves high detection accuracy under varying environmental conditions, including changes in lighting, occlusions, and traffic density. The modular design of the system allows for scalability and integration with existing smart city infrastructure. Additionally, the system can be extended to detect other traffic violations, making it a versatile solution for intelligent traffic management.In conclusion, this research contributes to the advancement of automated traffic surveillance by providing a reliable, scalable, and efficient solution for detecting critical safety violations. The deployment of such systems can significantly enhance road safety, reduce accident rates, and support law enforcement agencies in maintaining traffic discipline.