Publication: Diseño de sistema experto para búsqueda de plazas libres en un aparcamiento mediante aprendizaje automático y visión artificial
Loading...
Official URL
Full text at PDC
Publication Date
2019
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
Encontrar sitios libres en aparcamientos dedicados es un problema cotidiano al que se enfrentan los conductores. Debido a la falta de conocimiento de la situación exacta de las posiciones libres dentro de los aparcamientos, pierden una cantidad de tiempo innecesario en la búsqueda de huecos en los que poder aparcar. Un sistema capaz de estimar las características del vehículo a estacionar y de ubicar los huecos libres válidos para el tamaño de ese vehículo, disminuiría los tiempos de búsqueda empleados por los conductores y conseguiría un reparto óptimo de las plazas del aparcamiento.
El sistema propuesto combina técnicas de inteligencia artificial, para determinar la ocupación y distribución de los vehículos dentro del aparcamiento, y de entornos distribuidos, para optimizar los cómputos y escalabilidad del sistema. Mediante técnicas de visión artificial y redes neuronales el sistema implementa los módulos necesarios para la detección de vehículos dentro de un aparcamiento. Primero, el sistema genera un mapa computacional con las posiciones y tamaños de las plazas de aparcamientos. Segundo, identifica las características de los vehículos entrantes empleando técnicas de visión por computador para establecer las medidas. Tercero, identifica y diferencia los vehículos de otros objetos dentro del aparcamiento para conocer su ubicación exacta y tamaño empleando redes neuronales. Finalmente, el sistema identifica a los vehículos entrantes asignándoles la ubicación óptima en función de sus características, gracias al conocimiento de la ocupación del aparcamiento en tiempo real.
Se evalúa la precisión del sistema obteniendo resultados en la de detección de la ocupación del 93,39%, lo que indica el gran potencial del sistema como mejora a los enfoques actuales basados en la detección de aparcamiento en recintos exteriores mediante inteligencia artificial, proponiendo nuevos enfoques y características no estudiadas hasta el momento.
Finding free places in dedicated parking lot is a daily problem faced by drivers. Due to the lack of knowledge of the exact location of free positions within the car parks, they waste an unnecessary amount of time searching for parking space. A system capable of estimating the characteristics of the vehicle to be parked and of locating the free spaces valid for the size of that vehicle would reduce the search times used by drivers and achieve an optimum distribution of parking spaces. The proposed system combines artificial intelligence techniques, to determine the occupation and distribution of vehicles within the car park, and distributed environments, to optimize the computations and scalability of the system. By means of artificial vision techniques and neural networks, the system implements the necessary modules for the detection of vehicles inside a car park. First, the system generates a computer map with the positions and sizes of the parking spaces. Second, it identifies the characteristics of incoming vehicles using computer vision techniques to establish the measurements. Third, it identifies and differentiates vehicles from other objects within the car park in order to know their exact location and size using neural networks. Finally, the system identifies incoming vehicles by assigning them the optimum location according to their characteristics, thanks to the knowledge of the occupation of the car park in real time. The precision of the system is evaluated, obtaining results in the detection of occupancy of 93.39%, which indicates the great potential of the system as an improvement to the current approaches based on the detection of parking in outdoor areas by artificial intelligence, proposing new approaches and features not studied so far.
Finding free places in dedicated parking lot is a daily problem faced by drivers. Due to the lack of knowledge of the exact location of free positions within the car parks, they waste an unnecessary amount of time searching for parking space. A system capable of estimating the characteristics of the vehicle to be parked and of locating the free spaces valid for the size of that vehicle would reduce the search times used by drivers and achieve an optimum distribution of parking spaces. The proposed system combines artificial intelligence techniques, to determine the occupation and distribution of vehicles within the car park, and distributed environments, to optimize the computations and scalability of the system. By means of artificial vision techniques and neural networks, the system implements the necessary modules for the detection of vehicles inside a car park. First, the system generates a computer map with the positions and sizes of the parking spaces. Second, it identifies the characteristics of incoming vehicles using computer vision techniques to establish the measurements. Third, it identifies and differentiates vehicles from other objects within the car park in order to know their exact location and size using neural networks. Finally, the system identifies incoming vehicles by assigning them the optimum location according to their characteristics, thanks to the knowledge of the occupation of the car park in real time. The precision of the system is evaluated, obtaining results in the detection of occupancy of 93.39%, which indicates the great potential of the system as an improvement to the current approaches based on the detection of parking in outdoor areas by artificial intelligence, proposing new approaches and features not studied so far.
Description
Trabajo de Fin de Máster, Universidad Complutense, Facultad de Informática, Departamento de Arquitectura de Computadores y Automática, Curso 2018/2019