Helmet Detection and Number Plate Extraction System for Automated Traffic Violation Monitoring

Two-wheelers are the most common modes of transportation in India. Traffic accidents are the major cause of death globally, particularly in India. Although two-wheeled motorcycles are a natural alternative for a practical mode of transportation, they also significantly increase the number of fatalities and injuries in traffic accidents. Despite government regulations, many people opt not to wear helmets while driving. According to government data, more than 10,000 people die in motorcycle accidents in India each year, and a considerable fraction of these deaths might be avoided if riders wore helmets. The most popular method for ensuring that motorcyclists wear helmets is for traffic officers to manually check motorcyclists at traffic junctions or through CCTV video and penalize those who do not wear a helmet. However, it needs human interaction and effort. The goal of this study is to create a robust system for helmet recognition and number plate extraction from motorcycle photos and videos utilizing the YOLO (You Only Look Once) object detection method and OpenCV-based Optical Character Recognition (OCR) technology. The method is intended to address the growing number of accidents caused by riders who do not wear helmets, as well as the necessity for improved automobile traffic management. To detect the presence of helmets on motorcyclists, the YOLO algorithm is trained on a huge dataset of motorcycle photos. This method provides an automated approach for recognizing non-helmeted motorcyclists and a system for collecting motorcycle license plates from CCTV camera footage. To begin, the system categorizes moving objects as either motorcycling or non-motorcycling using YOLO and then it detects and localize the presence of helmet whether the rider is wearing or not. Finally, the rider who was not wearing a helmet is recognized. If the quality of footage is low, we will use Laplacian filtration method to filter the video and minimise the excess noise from it. Then the number plate characters are extracted using OCR technology. The system is tested on a large set of real-world images and videos, and the results demonstrate high accuracy in detecting helmets and extracting number plate information.