Zoom control to compensate camera translation within a robot egomotion estimation approach

We previously proposed a method to estimate robot egomotion from the deformation of a contour in the images acquired by a robot-mounted camera [2, 1]. The fact that the contour should always be viewed under weak-perspective conditions limits the applicability of the method. In this paper, we overcom...

ver descrição completa

Detalhes bibliográficos
Autores: Alenyà Ribas, Guillem|||0000-0002-6018-154X, Torras, Carme|||0000-0002-2933-398X
Tipo de documento: capítulo de livro
Data de publicação:2007
País:España
Recursos:Universitat Politècnica de Catalunya (UPC)
Repositório:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglês
OAI Identifier:oai:upcommons.upc.edu:2117/2673
Acesso em linha:https://hdl.handle.net/2117/2673
Access Level:Acceso aberto
Palavra-chave:Computer vision
Visió per ordinador
Classificació INSPEC::Pattern recognition::Computer vision
Àrees temàtiques de la UPC::Enginyeria de la telecomunicació::Processament del senyal::Processament de la imatge i del senyal vídeo
Descrição
Resumo:We previously proposed a method to estimate robot egomotion from the deformation of a contour in the images acquired by a robot-mounted camera [2, 1]. The fact that the contour should always be viewed under weak-perspective conditions limits the applicability of the method. In this paper, we overcome this limitation by controlling the zoom so as to compensate for robot translation along the optic axis. Our control entails minimizing an error signal derived directly from image measurements, without requiring any 3D information. Moreover, contrarily to other 2D control approaches, no point correspondences are needed, since a parametric measure of contour deformation suffices. As a further advantage, the error signal is obtained as a byproduct of egomotion estimation and, therefore, it does not introduce any burden in the computation. Experimental results validate this zooming extension to the method. Moreover, robot translations are correctly computed, including those along the optic axis.