Fusión multicámarapara seguimiento de objetos basada en teoría evidencial

Using multiple cameras has proven to increase vision systems capabilities, mainly by extending visual field or complementing information from cameras to reduce uncertainty. While there are emerging approaches which take in consideration information such as where an object is standing or how long it...

Descripción completa

Detalles Bibliográficos
Autor: ESTEBAN OMAR GARCIA RODRIGUEZ
Tipo de recurso: tesis de maestría
Estado:Versión aceptada para publicación
Fecha de publicación:2008
País:México
Institución:Instituto Nacional de Astrofísica, Óptica y Electrónica
Repositorio:Repositorio Institucional del INAOE
Idioma:español
OAI Identifier:oai:inaoe.repositorioinstitucional.mx:1009/440
Acceso en línea:http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/440
Access Level:acceso abierto
Palabra clave:info:eu-repo/classification/Cámaras/Cameras
info:eu-repo/classification/Rastreo/Tracking
info:eu-repo/classification/Fusión/Fusion
info:eu-repo/classification/cti/1
info:eu-repo/classification/cti/12
info:eu-repo/classification/cti/1203
Descripción
Sumario:Using multiple cameras has proven to increase vision systems capabilities, mainly by extending visual field or complementing information from cameras to reduce uncertainty. While there are emerging approaches which take in consideration information such as where an object is standing or how long it has been on a particular zone, it is still no common to find work related to the use of multiple cameras to improve information useful for such surveillance systems. In this work a decision fusion level model is proposed to take advantage of several geographically distributed cameras, in order to reduce uncertainty derived from cameras’perspective. Previously defined zones are considered to track objects position, so the stage of the processing at which data integration takes place is useful for high level surveillance systems, such those focused on behavior recognition. In our model, individual decisions are taken by means of an axisprojection- based generalized basic belief assignment (gbba) function and finallyfused using Dezert-Smarandache (DSm) hybrid rule. It is also proposed a way to reduce and manage dynamically the frame of discernment to optimize computer resources. Results of model are presented, obtained from tests on animated simulations and real sequences, and compared to a bayesian fusion model. Experiments proved that the proposed model yields a good improvement in tracking accuracy at high level processing.