Helmet Detection and Number Plate Recognition for Safety and Surveillance System

Abstract

With bikes becoming a less expensive and normal mode of transportation, the range of bike accidents is increasing unexpectedly as maximum motorcyclists are not wearing helmets, making every experience on a bike risky regularly. Even though the present CCTV-based systems are powerful, they require an excellent deal of human help, their efficiency declines through the years, and human bias is also a problem. Automation of this method is consequently fantastically ideal. In recent years, traffic accidents have increased sharply, and many people have been seriously injured or killed because of driving without wearing a motorcycle helmet. Not wearing a helmet is declared as a criminal offense by the government but even then, many people do not follow it. Traffic cops cannot arrest all violators at once, especially when traffic is heavy. In this study, we have proposed a system that can identify cyclists without helmets automatically and obtain information about their bike owners by recognizing license plates. The objects from video frames are identified using Object Detection API named TensorFlow. The proposed model is trained to recognize helmet-less cyclists using a faster R-CNN. A Tesseract OCR engine is then used to detect the license plate. Whenever any Helmetless rider is found or detected then the License Plate is fetched and then the number of that License Plate is identified by the use of an Optical Character Recognizer. In real-time, this system can be used by using a Web camera or a CCTV for providing input to the system. Experimental results of this model will show better performance than the prior art.