An AI-Enabled Traffic Monitoring System for Helmet Violation Detection and License Plate Recognition

Abstract

The project combines object detection and optical character recognition (OCR) algorithms to automatically detect helmet usage and read vehicle number plates in real-time from video streams. The system employs a pre-trained YOLO model for identifying key objects such as persons, helmets, motorbikes, and number plates. PaddleOCR is utilized to extract textual information from the detected number plates. Helmet compliance is evaluated by analyzing the spatial relationship between detected individuals and helmets. If a person is identified without a helmet and associated with a motorcycle and a readable number plate, the system logs the number plate and helmet status. The results are visually annotated on the video stream and stored in an Excel file for reporting and enforcement purposes. This system can enhance road safety surveillance and automate the detection of traffic rule violations.