[AllUsers-ISR] Palestras em tópicos de Visão por Computador

[Helder Araujo]

Olá a todos,

Durante a próxima semana haverá duas palestras relativas a tópicos
de Visão por Computador. Em ambas as palestras se discutem as
aplicações da estimação de rectas 3D a partir de dados como
câmaras, LIDARs e sensores equivalentes.

A 1ª palestra terá lugar no Anfiteatro do ISR, às 11 horas do dia
23 de Maio e terá como tópico "Análise de Rectas em Superfícies
com Simetria Rotacional. Título e bio:

________________________________________________

_*Calculus of lines on surfaces with rotational symmetry*_

In this talk we focus on 3D-surfaces that are swept by a moving line

that rotates around a fixed axis. These surfaces appear in CAD and 
engineering
applications with Galvano laser scanners, Lidars, rotating camera's,
moving robot joint axes, etc...

We explain how we can represent a ruled surface of revolution by 3 lines,
such that each of its lines is obtained as a linear combination of this 
triple.
Furthermore, the coefficients of these linear combinations can be
directly expressed in function of the rotation angles.

As an application we briefly show how these linear combinations
provide an alternative calibration model for certain 3D-sensors,
avoiding the determination of intrinsic parameters.

_*Bio:*_

Rudi Penne has been trained as a mathematician at the
University of Antwerp (Belgium), where he received his doctorate in 1992,
in the twilight zone between algebra, combinatorial and projective 
geometry,
and mechanical applications. His current position is a logical
continuation of his PhD research: professor at the
faculty of applied engineering of the University of Antwerp.

His main fields of interest: computer vision (geometric modelling
and calibration), projective (line) geometry, rigidity theory and
combinatorial/computational geometry).

A significant portion of his time is dedicated to the popularization
and communication of maths to a broad public.

______________________________________________________________________

  A segunda "talk" terá lugar também no Anfiteatro do ISR, às 14h 30m do dia
24 de Maio e terá como tópico a relação entre estimação de rectas em 3D e
"machine learning".

Título e bio:

__________________________________________________________


_*On straight lines and how machine learning screws them*_


Straight lines appear in a wide variety of industrial applications and
measuring techniques, such as robotics, LiDAR, cameras, 3D laser
scanners, … . After providing a few examples, we review the basics
of line geometry: Plücker coordinates for straight lines, the Klein
quadric and a little bit of screw theory. Armed with this knowledge,
we revisit the earlier examples. In the second half, we present challenges
we face when we apply constrained Gaussian process regression to predict
straight lines and not just numbers (the same goes for any machine learning
technique in a generative setting). We look at two use cases: the 
calibration
of a galvanometric laser setup and the approximation of point clouds using
line elements. This talk covers a large range of theoretical concepts
in an intuitive and visual way.

_*Bio*_

Ivan De Boi received his master’s degree in electronics and ICT
from the University of Antwerp, Antwerp, Belgium, in 2001. After working 
as a
Project Engineer for seven years in industry, he became a Lecturer at 
the Karel
de Grote University College, Antwerp. Apart from his teaching 
assignment, he also
worked on various research projects in the field of artificial 
intelligence and AR/VR,
mainly in healthcare and Industry 4.0 context. In 2020, he joined the
Research Group InViLab, University of Antwerp, to work on calibration
procedures of 3D sensors using line variety models.

______________________________________________________________

Cordiais saudações académicas,

Helder Araújo


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.isr.uc.pt/archive/allusers/attachments/20230517/e5f34098/attachment.html>