Real time license plate number extraction of non-helmet person using YOLO algorithm

Abstract

One of the problems in traffic regulations in India is riding motorcycle/mopeds without helmet, which increases accident sand deaths. In the existing system, the traffic police monitor the traffic violations through CCTV recordings, and in case if the rider without helmet is detected, then its vehicle number is recorded. But the constant monitoring is required to control the traffic rule violation which happens very frequently. To overcome these problems, we will require a system which would automatically handle traffic violations for non-helmet rider and thus would automatically extract the vehicles’ license plate number. The various research has successfully done in this area using CNN, R-CNN, LBP, HoG, HaaR features etc., but the results are limited with respect to efficiency, accuracy and speed. To overcome the problems associated with it, we develop a Non-Helmet Rider detection system, which attempts to satisfy the automation of detecting the traffic violation of non-helmet person and extracting the vehicles’ license plate number. The main principle involved in this system is Object Detection using Deep Learning at three levels. The person, motorcycle/moped is detected at first level using YOLOv2, helmet at second level using YOLOv3, License plate at the last levelusing YOLOv2.