DEEP LEARNING-BASED SYSTEM FOR NON-HELMET RIDER DETECTION AND LICENSE PLATE RECOGNITION TO ENHANCE ROAD SAFETY

Abstract

It is essential to have efficient automatic helmet detection in order to improve the safety of drivers on the road. On the other hand, the automated systems that are now in use frequently fall short in terms of efficiency, accuracy, and speed when it comes to object recognition and classification. This study presents a Non-Helmet Rider detection system with the objective of automating the identification of traffic offenses linked to helmet non-compliance. Additionally, the system is designed to retrieve the license plate numbers of automobiles. Object Detection by Deep Learning is the fundamental principle that underpins this system. This deep learning system functions on three different levels. Individuals, motorbikes and mopeds (at the first level), helmets (at the second level), and license plates (at the third level) are all detected by the system. YOLOv2 allows for the detection of license plates. After that, the number that is printed on the license plate is extracted using a technique called optical character recognition (OCR). The application of all of these strategies is carried out while taking into account the conditions and constraints that have been set, with particular emphasis paid to the process of extracting license plate numbers. As a result of the fact that this system processes image inputs, the speed of execution is of essential significance.