| |
|
Description: |
Central cameras are vision systems with a single effective
viewpoint. In this cameras the light rays forming the image intersect in
a single point in space called the projection center.
Examples of central projection systems are the conventional
perspective camera, cameras with radial distortion
(wide-angular lens), and catadioptric sensors (arrangements of
lenses and mirrors). The image geometry of perspective cameras
in the context of self-calibration and structure from motion
is well known and subject of textbooks. However, such theory
does not apply to central projection cameras other than the pin-hole.
In this work we show how to extend the theory for the cases of
cameras with wide-angle lenses and catadioptric sensors. We
propose an embedding of the image plane
in the 5D projective space in order to deal with the non-linearities
of the projection model. The projection of lines is discussed and
it is shown how the sensors can be easily calibrated using line images.
Additionally, we prove the existence of a general fundamental
matrix relating corresponding points in pairs of views acquired by any
mixture of central cameras. A linear algorithm to estimate the
fundamental matrix for images with radial distortion is proposed.
The usefulness of the framework is illustrated in the calibration of
a wide-area camera network.
Area(s):
|
|
Date: |
|
| Start Time: |
14:30 |
|
Speaker: |
João Pedro Barreto (Dep. Engenharia Electrotécnica e de Computadores, FCTUC)
|
|
Place: |
Room 5.5
|
| Research Groups: |
-Numerical Analysis and Optimization
|
|
See more:
|
<Main>
|
|
