Citas bibligráficas
Calle, R., (2019). Reconocimiento de placas vehiculares usando redes neuronales y un sistema de super resolución de imágenes basado en métodos Kernel usando GPUs [Tesis, Universidad Nacional de Ingeniería]. http://hdl.handle.net/20.500.14076/18492
Calle, R., Reconocimiento de placas vehiculares usando redes neuronales y un sistema de super resolución de imágenes basado en métodos Kernel usando GPUs [Tesis]. PE: Universidad Nacional de Ingeniería; 2019. http://hdl.handle.net/20.500.14076/18492
@misc{renati/710318,
title = "Reconocimiento de placas vehiculares usando redes neuronales y un sistema de super resolución de imágenes basado en métodos Kernel usando GPUs",
author = "Calle Flores, René Eduardo",
publisher = "Universidad Nacional de Ingeniería",
year = "2019"
}
The present work describes the development of an automatic number plate recognition system using neural networks and a super-resolution system in order to improve the accuracy of the ANPR system. The system is designed to detect Peruvian plates that meet the characteristics established in the national plate regulations. The proposed system consists of stages or modules that perform a specific task taking as input the images generated by digital cameras or video cameras. The first stage is to locate the possible plates present in images of an established size of 1280x720 pixels. In the second stage, a higher resolution image is generated from each image of each of the plates found, and using image processing methods to extract possible characters. Finally, in the character recognition stage, characters are normalized by adjusting them to a standard size of 22x36, and then use a 3-layer Multilayer Neural Network. The prototype system implemented has been successfully tested using images with various types of plates. It should be noted that our recognition system has a processing time of less than 300 milliseconds, and has been implemented in the high-level programming languages Python and CUDA C / C ++.
Este ítem está sujeto a una licencia Creative Commons Licencia Creative Commons