Banca de QUALIFICAÇÃO: ELIZABETH VIVIANA CABRERA AVILA

Uma banca de QUALIFICAÇÃO de DOUTORADO foi cadastrada pelo programa.
DISCENTE : ELIZABETH VIVIANA CABRERA AVILA
DATA : 04/12/2018
HORA: 10:30
LOCAL: Sala 02-DCA
TÍTULO:

Fusion of Stereo Pairs of Monocular Visual Odometry by Kalman Filtering


PALAVRAS-CHAVES:

visual odometry, fusion, Kalman filter, error caracterization, perturbation theory, Polynomial Eigenvalues Problem


PÁGINAS: 50
GRANDE ÁREA: Ciências Exatas e da Terra
ÁREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
ESPECIALIDADE: Processamento Gráfico (Graphics)
RESUMO:

Monocular visual odometry is a process that incrementally estimates the pose of a robotic agent from the analysis of images captured by a camera. This procedure can be the core of a visual navigation system or be part of more complex systems, including optical sensors coupled to wheels, GPS, inertial measurement units (IMUs), laser odometry, etc. There is a growing demand for real-time high-precision navigation systems (which in one form or another include visual information), which leads to the search for versatile and scalable solutions. In this context, our work proposes to take advantage of the monocular visual odometry and to improve its performance (precision and time) through a statistical fusion with the Kalman Filter of two odometry obtained from the individual captures of a stereo camera. Each of the estimates will be evaluated on the basis of their errors, based on the propagation of perturbations in eigenvalues applied in the algorithm of estimation of the essential matrix of five points. To overcome the numerical errors evidenced in the calculation of visual odometry algorithms and reduce their processing time, the goal is to include a mathematical framework based on the Polynomial Eigenvalues Problem (PEP).


MEMBROS DA BANCA:
Presidente - 1345674 - LUIZ MARCOS GARCIA GONCALVES
Interno - 2929823 - RAFAEL BESERRA GOMES
Externo ao Programa - 1170845 - BRUNO MARQUES FERREIRA DA SILVA
Externo à Instituição - ANDERSON ABNER DE SANTANA SOUZA - UERN
Notícia cadastrada em: 20/11/2018 15:48
SIGAA | Superintendência de Tecnologia da Informação - (84) 3342 2210 | Copyright © 2006-2024 - UFRN - sigaa05-producao.info.ufrn.br.sigaa05-producao