Real-Time Surveillance For Road Safety: Object Detection and Compliance Monitoring

Assistant Professor Dr. Pon Partheeban, Aslin R, Harish Suresh Kumar, Roshan Lal J, Vijesh G

Abstract- – Road safety is a critical concern worldwide, with two-wheeler violations posing significant risks to riders and other road users. This project presents a comprehensive real- time surveillance system for monitoring and detecting road safety violations by two-wheelers using state-of-the-art computer vision techniques. The proposed approach leverages the powerful YOLOv8 object detection architecture to accurately identify and localize two-wheelers, riders, helmets, mobile devices, and license plates in video streams. The system employs a multi- task learning strategy, where a single deep neural network is trained to simultaneously detect and classify multiple objects of interest. Post-processing algorithms analyze the spatial and contextual relationships between detected objects to identify specific violations, such as lack of helmet usage, triple riding, and distracted driving due to mobile phone usage. To ensure real-world applicability, the system is designed for seamless integration with existing traffic monitoring infrastructure, enabling real-time violation detection and automated reporting to relevant authorities. Extensive experiments on diverse datasets demonstrate the system’s robustness and efficiency, achieving state-of-the-art performance in detecting road safety violations by two-wheelers. The proposed solution offers a scalable and cost- effective approach to improving road safety, with the potential to significantly reduce accidents and casualties involving two- wheelers. This project contributes to the field of computer vision and intelligent transportation systems by presenting a novel application of deep learning object detection for enhancing road safety compliance monitoring.

DOI: /10.61463/ijset.vol.12.issue2.694