Real-time license plate identification of moving vehicles using Deep Learning

Abstract

As the population in India keeps growing day by day, so does the traffic. At present, the backbone of the toll booth in India is the FASTag, which requires that the vehicles come to a complete halt. This results in clogging of traffic, which wastes both time and fuel. The growth of smart cities and smart transportation systems has led to the growing need of sustainable automatic license plate recognition systems. It is crucial that a system not only accurately identify the number plates, but also perform at high speeds to not disrupt the flow of traffic. This article presents an approach through Deep Learning techniques like Object Detection and Optical Character Recognition, which are purveyors of both speed and accuracy. The proposed License Plate Identification system is an end-to-end solution that works on moving vehicles in real time. The system has been tested thoroughly and the findings indicate that the pipeline can accurately predict the number plates of vehicles traveling at speeds of 35–40[Formula: see text]km/h. Further testing will be conducted for deployment on edge devices and GPUs.