Bibliographic citations
This is an automatically generated citacion. Modify it if you see fit
Nizama, D., (2022). Sistema de detección y clasificación vehicular basado en redes neuronales de aprendizaje profundo [Tesis, Universidad Nacional de Ingeniería]. http://hdl.handle.net/20.500.14076/23816
Nizama, D., Sistema de detección y clasificación vehicular basado en redes neuronales de aprendizaje profundo [Tesis]. PE: Universidad Nacional de Ingeniería; 2022. http://hdl.handle.net/20.500.14076/23816
@misc{renati/712383,
title = "Sistema de detección y clasificación vehicular basado en redes neuronales de aprendizaje profundo",
author = "Nizama Yamunaque, David Alexander",
publisher = "Universidad Nacional de Ingeniería",
year = "2022"
}
Title: Sistema de detección y clasificación vehicular basado en redes neuronales de aprendizaje profundo
Authors(s): Nizama Yamunaque, David Alexander
Advisor(s): Flores González, Leonardo
Keywords: Automatización; Sistema vehicular; Transporte vehicular; Ingeniería del tránsito; Redes neuronales
OCDE field: http://purl.org/pe-repo/ocde/ford#2.01.01
Issue Date: 2022
Institution: Universidad Nacional de Ingeniería
Abstract: La presente tesis presenta un sistema de detección, clasificación y conteo vehicular mediante el uso de redes neuronales de aprendizaje profundo y el de la inteligencia artificial, específicamente del área de visión artificial, con el objetivo de realizar conteos y aforos vehiculares. Para ello, se utilizarán diversos algoritmos y redes neuronales previamente entrenadas que permiten dotar al ordenador la capacidad de ver y comprender el contenido de imágenes y videos a través del reconocimiento de patrones y características.
El aforo vehicular es uno de los aspectos más importantes y primarios en un estudio de tráfico vehicular, debido a que a partir de estos datos se determina el grado de ocupación y condiciones en las que una vía funciona, así como, las futuras tendencias de crecimiento, lo que permite una correcta planificación y diseño de una construcción, rehabilitación o mejora de una obra vial. Por ello, la finalidad del presente trabajo de investigación es brindar una alternativa accesible, rentable y económica que permita realizar aforos vehiculares en una vía mediante las virtudes de la inteligencia artificial, las cuales en los últimos años han tenido un desarrollo y progreso destacable.
Para la detección y clasificación de vehículos se utilizaron las redes neuronales convolucionales, las cuales están diseñadas para imitar la corteza visual del cerebro y reconocer objetos en imágenes y videos. Estás redes contienen una serie de capas jerarquizadas y especializadas que permiten identificar y diferenciar un objeto de otro, por lo que fue factible clasificar los vehículos de acuerdo a su tipología, esto aportó a obtener una data completa y confiable.
Palabras claves: Inteligencia artificial, visión artificial, redes neuronales, aprendizaje profundo, aforo vehicular, estudio de tráfico vehicular, algoritmo de detección, reconocimiento de patrones, convoluciones.
This thesis presents a vehicle detection, classification and counting system through the use of deep learning neural networks and artificial intelligence, specifically in the area of artificial vision, with the objective of performing vehicle counting and gauging. For this purpose, several algorithms and previously trained neural networks will be used to provide the computer with the ability to see and understand the con-tent of images and videos through the recognition of patterns and features. The vehicle capacity is one of the most important and primary aspects in a study of vehicular traffic, because these data determine the degree of occupation and conditions in which a road works, as well as future growth trends, which allows a correct planning and design of a construction, rehabilitation or improvement of a road work. Therefore, the purpose of this research work is to provide an accessible, profitable and economical alternative that allows to perform vehicular gauging on a road through the virtues of artificial intelligence, which in recent years have had a remarkable development and progress. For the detection and classification of vehicles, convolutional neural networks we-re used, which are designed to imitate the visual cortex of the brain and recognize objects in images and videos. These networks contain a series of hierarchical and specialized layers that allow to identify and differentiate one object from another, so it was feasible to classify vehicles according to their typology, this contributed to obtain a complete and reliable data.
This thesis presents a vehicle detection, classification and counting system through the use of deep learning neural networks and artificial intelligence, specifically in the area of artificial vision, with the objective of performing vehicle counting and gauging. For this purpose, several algorithms and previously trained neural networks will be used to provide the computer with the ability to see and understand the con-tent of images and videos through the recognition of patterns and features. The vehicle capacity is one of the most important and primary aspects in a study of vehicular traffic, because these data determine the degree of occupation and conditions in which a road works, as well as future growth trends, which allows a correct planning and design of a construction, rehabilitation or improvement of a road work. Therefore, the purpose of this research work is to provide an accessible, profitable and economical alternative that allows to perform vehicular gauging on a road through the virtues of artificial intelligence, which in recent years have had a remarkable development and progress. For the detection and classification of vehicles, convolutional neural networks we-re used, which are designed to imitate the visual cortex of the brain and recognize objects in images and videos. These networks contain a series of hierarchical and specialized layers that allow to identify and differentiate one object from another, so it was feasible to classify vehicles according to their typology, this contributed to obtain a complete and reliable data.
Link to repository: http://hdl.handle.net/20.500.14076/23816
Discipline: Ingeniería Civil
Grade or title grantor: Universidad Nacional de Ingeniería. Facultad de Ingeniería Civil
Grade or title: Ingeniero Civil
Juror: Santa María Dávila, Edward; Matías León, José Carlos
Register date: 9-Mar-2023
This item is licensed under a Creative Commons License