SILVA, V. C.; http://lattes.cnpq.br/4668405185898453; SILVA, Vânia Cordeiro da.
Resumo:
The objective of this dissertation is the development of a method for
identification of 3D vertices for multi-view images of scene composed by one or more
objects. This identification allows the definition of polyhedral object models, which
can be used for 3D recognition. The developed algorithms operate by data-flow, thus
enabling the whole modelling process to be perfomed in real-time. The method works
with three real images of a scene, taken from three different points of view, also known
as multi-view images: top, side and front views. The method recognizes not only the
external contour of the object, but also its internal edges.