Extracting the Number from License plate of Non Helmet Riders

Abstract

In India's present scenario, there are many issues with traffic laws that may be addressed using a variety of approaches. A traffic infraction that has led to a rise in the number of accidents and fatalities in India is riding a motorbike or scooter without a helmet. In the current system, traffic infractions are mainly monitored via CCTV footage, where traffic police must look into the frame where the traffic violation occurs, zoom into the licence plate if the rider is not wearing a helmet, and so forth. The problem is that traffic infractions are becoming more common, and the number of individuals riding motorbikes is growing at an alarming rate. What if there was a system that looked for traffic violations such as not wearing a helmet while riding a motorbike or moped and, if any were found, automatically extracted the licence plate number of the vehicle in question. Recent studies using CNN, R-CNN, LBP, HoG, and HaaR characteristics, among others, have been effective in accomplishing this task. There are a few drawbacks to these methods, including their inability to identify and classify objects quickly or accurately. Non-Helmet Rider detection system is developed in this study effort to fulfil the automation of traffic infraction detection and licence plate number extraction. Deep Learning is used at three layers for Object Detection in this project. YOLOv2 detects a person, motorcycle/moped, helmet, and licence plate at the first level. YOLOv2 detects a licence plate at the second level. After that, OCR is used to retrieve the licence plate number (Optical Character Recognition). All of these methods, particularly the one that involves obtaining a licence plate number, are governed by predetermined rules and regulations. Since this project uses video as an input, it is critical that it be completed quickly. This methodology has been utilised to create an integrated system that can identify helmets and retrieve licence plate numbers.