Deep Learning based Traffic Infringement Recognition System

Abstract

The majority of developing nations in the emerging global South now face problems related to traffic law violations. There has been a dramatic increase in both the number of cars on the road and the number of infractions of traffic laws. Traffic infraction management has long been a difficult and perhaps risky profession. Although traffic management has become automated, it still poses a difficult task owing to the wide variety of Plate formats, scales, rotations, and lighting conditions encountered during picture acquisition. This project's primary goal is to efficiently and affordably curb infractions of traffic laws. The suggested concept makes use of an automated camera system to record video and take still images. Automatic Number Plate Recognition (ANPR) and other image processing algorithms for plate localisation and character recognition are presented in this work, facilitating the rapid and accurate identification of both vehicles and their corresponding licence plates. The system works by first identifying the offending vehicle's number plate, and then using an SMS-based system to alert the vehicle's owners about the infraction. When a vehicle's report is submitted, the Regional Transport Office (RTO) receives an extra text message so that it can monitor the progress of the submission.