Publication: Segmentación de campo de juego en vídeos de baloncesto para detección de eventos en tiempo real
Loading...
Identifiers
Publication date
2010-09
Defense date
2010-09-23
Authors
Advisors
Tutors
Journal Title
Journal ISSN
Volume Title
Publisher
Impact
Export
Abstract
Las retransmisiones de vídeos deportivos, debido al gran atractivo que despiertan en una numerosa audiencia y a los importantes intereses económicos que existen en torno a ellas, llevan años siendo un importante motor del desarrollo y promoción de tecnologías tanto de comunicación como de procesado de imágenes. En este contexto existe un creciente interés por ser capaz de extraer características de secuencias de vídeos deportivos que permitan la detección de eventos del juego durante su retransmisión (en tiempo real) para enriquecer la información que se le proporciona al telespectador. Este proyecto consiste en el estudio, diseño e implementación de una parte de un sistema de este tipo: un segmentador del campo de baloncesto. Nuestro objetivo es separar el campo de juego en secuencias de vídeo de baloncesto FIBA de manera robusta, extensible a cualquier campo de baloncesto, y e ciente, para poder integrarla en un sistema de detección de eventos. En primer lugar se estudió el estado del arte en segmentación de campo, no sólo en vídeos de baloncesto sino también en otros deportes. A partir de este estudio se diseñaron e implementaron diversos prototipos evolutivos del sistema. De todos ellos se analizó su funcionalidad con sendos bancos de pruebas extraídos de vídeos de varios partidos. Finalmente se realizaron pruebas de rendimiento que nos permitieran evaluar la versión nal en términos de robustez y e ciencia. Aún sin cumplir plenamente los objetivos propuestos, los resultados obtenidos llevan al optimismo y nos permiten abrir nuevas líneas de investigación en este campo. _____________________________________________________________________________________
Sport video broadcasts, as they appeal to large audiences and due to the number of economic assets involved around them, have been an important development and promotion driving force of both communication and image processing technologies. In this context there is a growing interest in being able to extract features of sport video sequences that allow the detection of game events during the broadcast (in real-time) to enhance to information delivered to the viewer. This project consists in the study, design and implementation of a part of this type of system: a basketball court segmentation system. Our objective is to segment the court in FIBA basketball video sequences in a robust, it must be extended to any basketball court, and e cient way, so we can integrate it in a event detection system. Firstly the state of the art of court segmentation was studied, not only of basketball videos but also of other sports. Based on this study some evolutive prototypes of the system were design and implemented. The Functionality of all of them was studied with some test benches extracted from a number of game videos. Finally, some performance tests were performed in order to evaluate the last prototype in terms of robustness and e ciency. Although the stated objectives were not completely achieved, the obtained results give us reasons to be optimistic and allow us to open new research topics on this area.
Sport video broadcasts, as they appeal to large audiences and due to the number of economic assets involved around them, have been an important development and promotion driving force of both communication and image processing technologies. In this context there is a growing interest in being able to extract features of sport video sequences that allow the detection of game events during the broadcast (in real-time) to enhance to information delivered to the viewer. This project consists in the study, design and implementation of a part of this type of system: a basketball court segmentation system. Our objective is to segment the court in FIBA basketball video sequences in a robust, it must be extended to any basketball court, and e cient way, so we can integrate it in a event detection system. Firstly the state of the art of court segmentation was studied, not only of basketball videos but also of other sports. Based on this study some evolutive prototypes of the system were design and implemented. The Functionality of all of them was studied with some test benches extracted from a number of game videos. Finally, some performance tests were performed in order to evaluate the last prototype in terms of robustness and e ciency. Although the stated objectives were not completely achieved, the obtained results give us reasons to be optimistic and allow us to open new research topics on this area.
Description
Keywords
Proceso de imágenes, Video, Retransmisiones deportivas