Desarrollo de un prototipo para automatizar el registro de acceso vehicular mediante reconocimiento de placas utilizando redes neuronales.

Abstract

One of the challenges in residential buildings and commercial stores is the correct management of vehicle registration, which is usually done manually. This causes errors in the correct recording of license plates. To address this problem, it is proposed the development of a prototype that automates vehicle entry and exit registration through license plate and character recognition using neural networks and computer vision. This system will allow real-time image capture, identifying and processing license plate characters to record each access in a centralized database, reducing errors and improving security. This project uses a quantitative and descriptive approach, validating the prototype through performance metrics such as accuracy, completeness, and mAP. A license plate detection model was trained, achieving 94% performance in license plate identification, although 6% false negatives were identified.