Identification of Motorcycle Traffic Violations with Deep Learning Method

Abstract

This paper proposes a system that can detect vehicle license plates for motorcycle riders who violate traffic regulations using the YOLO algorithm. This detector will be placed at a traffic light junction. The system has two main video processes: helmet and rearview glass detection and license plate reading. Furthermore, a warning to that specific violator will be sounded through the speaker. The system's three main components are a camera, computing unit, and speaker. This system is built using the YOLO algorithm, Optical Character Recognition (OCR), and Text-to-Speech. For the system to meet real-time requirements, the video data captured by the webcam is sent to a computer device for image processing and identifying motorists who violate traffic without wearing a helmet. Suppose the driver is identified as committing a violation. The OCR system will extract the license plate into text form. Then, the Text-to-Speech system will produce sound output containing license plate information. The data obtained from the system testing shows that the value generated by the YOLO darknet system can detect all categories with an accuracy of 93%. The OCR system for reading the letters and numbers on the license plate has a 95% success rate. The Text-to-Speech system has an accuracy rate of 100%.