|
The 3D models
are used in many applications
such as computer games, movies,
medical, training simulators,
augmented reality, archaeology
and industrial inspection. The
3D models can be artificially
generated using any
off-the-shelf modeling software
such as Autodesk Maya. The
artificial 3D models are
sufficient for most of the
entertainment industry such as
computer games or movies.
However, most of the
applications such as archaeology
or industrial inspection demand
true representation of the real
complex objects. The true
representation of the real
complex objects can be achieved
by measuring real shape and
photometric properties using
computer vision techniques. This
process involves several steps
which require manual operations
as well as technical skills. In
addition to time required for
manual operations, the
processing time for large amount
of data is considerable.
Automating the 3D digitization
process can improve efficiency
and quality; while reducing
operator skills and cost at the
same time. Presently there is no
fully automatic system
available, although some
semi-automatic systems have been
developed.
The 3D
digitization process can be
divided into two general steps,
i.e. Acquisition and
Post-processing. The acquisition
system generally comprises a 3D
scanner mounted on a positioning
system. The post-processing step
includes registration, merging,
hole-filling, cleaning and
photometric and environment
mapping. The automation of
acquisition process requires
automatic path or view planning
in the presence of number of
constraints with regard to the
positioning system and the
scanner. The automation of
post-processing needs automation
of registration, merging,
hole-filling and cleaning etc.
There are many efficient
commercial solutions available
for the automation of most of
the post-processing steps, such
as
RapidForm from INUS
Technologies. However, there is
no commercial solution available
for the automation of 3D
acquisition process.
The crux of automatic
acquisition is the automatic
view planning. The view planning
has no or little a-prior
knowledge about the shape of the
object to be digitized most of
the time. However, some times
view planning may have a
complete or rough model of the
object to be digitized. There is
limited number of views
available subject to the
positioning system
involved. The purpose of this
project is to study and
implement methods for 3D
acquisition and post processing
to automatically generate 3D
models. Work concerns view
planning problem, views
registration, merging, holes
filling etc. |
