YOLOv8-Powered Multi-Object Detection for Safety Helmets and Number Plates in Real-Time Surveillance Systems

This paper presents a real-time object detection system using the latest YOLOv8 model to enhance traffic monitoring and safety enforcement. The model was trained to detect helmets, riders and number plates in diverse traffic scenarios, achieving an overall accuracy of 93.4%. Significant improvements were observed in precision, recall and mean Average Precision (mAP) scores due to the combination of an optimized architecture, high-quality data and fine-tuned hyperparameters. The system maintains real-time inference capabilities with a processing speed of 11 ms per image, making it suitable for continuous surveillance and enforcement applications. This study demonstrates the potential of using advanced deep learning techniques for effective and automated traffic rule enforcement. Future work may involve further optimizations and integration with large-scale cloud-based systems for more comprehensive deployments.