Real-Time Helmet and Triple Riding Violation Detection Using YOLOv8-Based Deep Learning and OCR Framework

Road accidents involving two-wheelers have become a major public safety concern due to increasing traffic violations such as riding without helmets and triple riding. Existing intelligent traffic monitoring systems based on earlier YOLO models, CNN frameworks, and traditional surveillance methods provide moderate detection performance; however, they still face several challenges including reduced accuracy under crowded traffic conditions, poor small-object detection, inconsistent object tracking, low OCR accuracy under varying illumination conditions, and limited capability for simultaneous multi-violation detection. To address these limitations, this paper proposes a real-time Helmet and Triple Riding Violation Detection System using the YOLOv8 deep learning framework integrated with CNN, DeepSORT tracking, and OCR technology. The proposed framework automatically detects motorcycles, riders, helmets, and multiple riders simultaneously from surveillance video streams and traffic images. YOLOv8 is utilized for high-speed object detection and accurate localization, while DeepSORT tracking maintains object identity consistency across continuous frames. OCR technology is incorporated for automatic vehicle registration number extraction and intelligent violation reporting. The proposed framework performs preprocessing operations including frame extraction, image resizing, normalization, and contrast enhancement before object detection. Experimental analysis demonstrates improved precision, recall, F1-score, OCR accuracy, and real-time performance compared with existing traffic monitoring approaches. The proposed intelligent surveillance framework supports automated traffic violation monitoring, reduces manual intervention, and improves smart transportation management efficiency.