Abstract: Nowadays, wearing a helmet is crucial for safety, providing essential head protection and reducing the risk of severe injuries, potentially saving lives for motorbike and bicycle riders. Conversely, not wearing a helmet poses serious risks, including increased vulnerability to head injuries, fatalities, and elevated accident susceptibility due to reduced visibility and non-compliance with traffic regulations. A real-time system can leverage deep learning to detect instances of helmet non-compliance. By employing the YOLOv3 algorithm, helmet non-compliance can be detected with an accuracy ranging between 80% and 95%. Furthermore, the XML framework is utilized for precise number plate extraction from vehicles of violators, ensuring similar accuracy levels. The EasyOCR algorithm converts these number plate images into text, facilitating their storage in a database.The system ensures comprehensive documentation and monitoring by securely storing the data of helmet non-compliance, including vehicle numbers, in a database. This supports detailed record-keeping, reporting, and further analysis, ultimately contributing to improved road safety and enforcement of helmet regulations.
Keywords: Deep learning, YOLOv3, XML framework, Easy OCR, Helmet non-compliance , Number plate extraction.
|
DOI:
10.17148/IJARCCE.2024.13830
