Automatic Detection of Helmetless Riders with License Plate Extraction

Helmet non-compliance among motorcyclists is a significant contributor to road traffic fatalities, particularly in densely populated regions. This study proposes a real-time, scalable framework for automatic detection of helmet violations and license plate recognition using the YOLOv8 object detection model. The system is designed to accurately distinguish between helmeted and non-helmeted riders, identify individual riders, and extract license plate details, even in complex traffic scenarios. Key enhancements include adaptive image preprocessing for improved visibility under poor lighting and fine-grained character recognition for accurate offender identification. The model was trained and validated on a diverse dataset and evaluated using standard metrics such as Precision, Recall, F1-score, mAP@50, and mAP@50:0.95. Results indicate high overall performance, with precision and recall exceeding 93 %, and mAP@ 50 of 93.89 %. Comparative analysis with state-of-the-art models such as SSD MobileNet V2 and Faster R-CNN Inception V2 confirms the superior performance of the proposed approach, particularly in rider detection. With an inference speed of 21.59 FPS, the system is suitable for integration with existing CCTV infrastructures for real-time law enforcement.