Use this link to cite:
http://hdl.handle.net/2183/31939 Análisis automático de flujo de vídeo CCTV para la estimación de ocupación de mercancías en entornos portuarios
Loading...
Identifiers
Publication date
Authors
Justo Armesto, Juan Andrés
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: La industria portuaria ofrece un elevado potencial para el desarrollo socioeconómico local
y regional por su situación estratégica para el intercambio de mercancías por vía marítima y
terrestre. Con el avance en las tecnologías de la información y las comunicaciones, los puertos
están avanzando hacia infraestructuras denominadas ”Smart Ports” o puertos inteligentes,
que son plataformas para la gestión global e inteligente que permiten la integración y comunicación
entre diferentes sistemas, facilitando la automatización de procedimientos complejos.
En este contexto, dado el volumen de tráfico y mercancías, las tareas de vigilancia de las
zonas de almacenamiento es crucial para garantizar la seguridad y la eficiencia en la operativa
diaria del puerto. La monitorización activa de las zonas de ocupación de superficies es una
tarea que consume tiempo y recursos, por lo que disponer de un sistema automático para este
proceso proporcionaría una herramienta útil y fiable de apoyo a la videovigilancia.
En este proyecto se propone un sistema automático basado en técnicas avanzadas de procesado
de imagen e inteligencia artificial para la estimación de los porcentajes de ocupación
de carga a partir de imágenes obtenidas por las cámaras Closed-Circuit Television (CCTV)
en zonas de interés definidas en instalaciones portuarias. Con este fin, por un lado se consideran
aproximaciones basadas en técnicas clásicas de aprendizaje máquina y por otro lado,
se propone el uso de técnicas basadas en deep learning. Para determinar qué método es más
apropiado para abordar este problema se ha realizado un estudio del impacto de diferentes
configuraciones con conjuntos de datos formado por imágenes capturadas por las cámaras
CCTV de diferentes instalaciones portuarias. Se ha realizado un análisis comparativo y se ha
seleccionado el método que mejores resultados proporciona para este dominio de aplicación.
Adicionalmente, el método seleccionado ha sido integrado en un servicio de videovigilancia
proporcionando una herramienta de apoyo que permite la monitorización fiable y objetiva de
las zonas de ocupación de superficie en un entorno real.
[Abstract]: The port industry offers a high potential for local and regional socio-economic development due to its strategic location for the exchange of goods by sea and land. With the advances in information and communication technologies, ports are moving towards infrastructures known as ”Smart Ports”, which are platforms for global and intelligent management that allow integration and communication between different systems, facilitating the automation of complex procedures. In this context, given the volume of traffic and goods, the surveillance of storage areas is crucial to guarantee security and efficiency in the port’s daily operations. The active monitoring of surface occupation zones is a time-consuming and resource-intensive task, so having an automatic system for this process would provide a useful and reliable tool to support video surveillance. This project proposes an automatic system based on advanced image processing and artificial intelligence techniques for estimating the percentages of cargo occupancy from images obtained by Closed-Circuit Television (CCTV) cameras in defined areas of interest in port facilities. To this end, on the one hand, approaches based on classical machine learning techniques are considered and, on the other hand, the use of techniques based on deep learning is proposed. In order to determine which method is more appropriate to address this problem, a study of the impact of different configurations with datasets composed of images captured by the CCTV cameras of different port facilities has been performed. A comparative analysis has been carried out and the method that provided the best results for this application domain has been selected. Additionally, the selected method has been integrated into a video surveillance service providing a support tool that allows the reliable and objective monitoring of surface occupation zones in a real environment.
[Abstract]: The port industry offers a high potential for local and regional socio-economic development due to its strategic location for the exchange of goods by sea and land. With the advances in information and communication technologies, ports are moving towards infrastructures known as ”Smart Ports”, which are platforms for global and intelligent management that allow integration and communication between different systems, facilitating the automation of complex procedures. In this context, given the volume of traffic and goods, the surveillance of storage areas is crucial to guarantee security and efficiency in the port’s daily operations. The active monitoring of surface occupation zones is a time-consuming and resource-intensive task, so having an automatic system for this process would provide a useful and reliable tool to support video surveillance. This project proposes an automatic system based on advanced image processing and artificial intelligence techniques for estimating the percentages of cargo occupancy from images obtained by Closed-Circuit Television (CCTV) cameras in defined areas of interest in port facilities. To this end, on the one hand, approaches based on classical machine learning techniques are considered and, on the other hand, the use of techniques based on deep learning is proposed. In order to determine which method is more appropriate to address this problem, a study of the impact of different configurations with datasets composed of images captured by the CCTV cameras of different port facilities has been performed. A comparative analysis has been carried out and the method that provided the best results for this application domain has been selected. Additionally, the selected method has been integrated into a video surveillance service providing a support tool that allows the reliable and objective monitoring of surface occupation zones in a real environment.
Description
Editor version
Rights
Atribución-NoComercial-SinDerivadas 3.0 España








