Page 234 - Innovations in Intelligent Machines

Toward Robot Perception through Omnidirectional Vision 227
Omnidirectional images can be generated by a number of different sys-
tems which can be classified into four distinct design groupings: Camera-Only
Systems; Multi-Camera – Multi-Mirror Systems; Single Camera – Multi-
Mirror Systems, and Single Camera – Single Mirror Systems.
Camera-Only Systems: A popular method used to generate omnidirectional
images is the rotation of a standard CCD camera about its vertical axis.
The captured information, i.e. perspective images (or vertical line scans) are
then stitched together so as to obtain panoramic 360 images. Cao et al.
◦
[11] describe such a system fitted with a fish-eye lens [60]. Instead of relying
upon a single rotating camera, a second camera-only design is to combine
cameras pointing in differing directions [28]. Here, images are acquired using
inexpensive board cameras and are again stitched together to form panoramas.
Finally, Greguss [40] developed a lens, he termed the Panoramic Annular Lens,
to capture a panoramic view of the environment.
Multi-Camera – Multi-Mirror Systems: This approach consists of arranging a
cluster of cameras in a certain manner along with an equal number of mirrors.
Nalwa [63] achieved this by placing four triangular planar mirrors side by side,
in the shape of a pyramid, with a camera under each. One significant prob-
lem with multi-camera – multi-mirror systems is geometric registering and
intensity blending the images together so as to form a seamless panoramic
view. This is a difficult problem to solve given that, even with careful align-
ment, unwanted visible artifacts are often found at image boundaries. These
occur not only because of variations between the intrinsic parameters of each
camera, but also because of imperfect mirror placement.
Single Camera – Multi-Mirror Systems: The main goal behind the design of
single camera – multi-mirror systems is compactness. Single camera – multi-
mirror systems are also known as Folded Catadioptric Cameras [66]. A simple
example of such a system is that of a planar mirror placed between a light ray
travelling from a curved mirror to a camera, thus “folding” the ray. Bruckstein
and Richardson [9] presented a design that used two parabolic mirrors, one
convex and the other concave. Nayar [66] used a more general design consisting
of any two mirrors with a conic-section profile.

Single Camera – Single Mirror Systems:

In recent years, this system design has become very popular; it is the approach
we chose for application to visual-based robot navigation. The basic method
is to point a CCD camera vertically up, towards a mirror.
There are a number of mirror profiles that can be used to project light
rays to the camera. The first, and by far the most popular design, uses a
standard mirror profile: planar, conical, elliptical, parabolic, hyperbolic
or spherical. All of the former, with obvious exception of the planar mirror,
◦
can image a 360 view of the environment horizontally and, depending on
the type of mirror used approximately 70 to 120 , vertically. Some of the
◦
◦