A deep-learning based real-time automatic ticketing system for high speed traffic

Abstract

This project focuses on enhancing the enforcement of traffic rules on roads by developing an automated ticketing system using deep learning. Despite law enforcement efforts, enforcing speed limits continues to pose a significant challenge. The main objective of this project is to design a real-time automatic ticketing system that simplifies the process of issuing speeding tickets, thereby improving the enforcement of traffic regulations and reducing violations. The system consists of several key components: vehicle detection, speed estimation, number plate recognition, and ticket generation. The vehicle detection module uses YOLO (You Only Look Once) to identify vehicles in video footage. Once a vehicle is detected, a speed estimation algorithm calculates its speed, and a number plate recognition system, employing character recognition, extracts the vehicle’s license plate number. This information is then used to generate and log speeding tickets automatically. Data for this project was collected along the Kampala-Hoima highway, Uganda, using cameras, phones, and speed guns to capture videos and snapshots of vehicles. The data was processed and then annotated. Two approaches were explored for number plate recognition: a transformer-based and a CNN-based approach.