Paper Contents
Abstract
This project presents a real-time traffic violation detection system that leverages deep learning and computer vision to identify license plates and helmet usage in video footage. The system integrates the YOLOv8 object detection model for recognizing license plates and detecting helmet presence, alongside EasyOCR for extracting alphanumeric text from detected plates. It incorporates a confidence-based OCR filtering mechanism and evaluates image clarity to detect potential tampering or unreadable plates. Violations such as missing license plates, low OCR confidence, invalid plate formats, and absence of helmets are automatically logged and stored with visual evidence. Additionally, the system computes performance metrics including precision, recall, and F1-score, and visualizes accuracy and loss over time. This solution aims to support intelligent surveillance systems and enhance traffic rule enforcement, particularly for two-wheeler riders.
Copyright
Copyright © 2025 TSKS Jyothirmayi. This is an open access article distributed under the Creative Commons Attribution License.
