|
State
of the art in Augmented Reality
Augmented Reality
(AR) stands for a supplementation of the real world rather than for a substitution.
It combines synthetic objects to real world, by superimposing computer-generated
information upon the real environment. To do so it makes use of the 3D
position and orientation of both the user and the scene. It allows to enhance
user’s performance within and beyond natural human perceptual channels.
Although the birth of AR goes back to the sixties, it took thirty years
to find applications outside research laboratories. The medical area was
one of the first areas of applications of AR. In this domain, AR provides
medical doctors with a kind of ”X-ray view” [Azuma: 97], [Hildebrand-2:
98]. Other applications have followed, such as repair of copier machines
[Feiner: 93], installation of aluminium struts [Webster et. al. 96], complex
maintenance and repair tasks [Azuma: 97], guided assembly [Sharma et. al:
97], electric wire bundle assembly in airplanes [Curtis et. al: 98], and
insertion of a lock into a car door [Reiners et. al 98]. Studies have been
made to use AR in the other areas, such as interior design [Siering: 97],
or special effects [CAS et. al.: 95].
However,
all these applications are somehow demonstrators and proof of concepts.
A real transfer to industrial use is still to be done. Since AR just started
to move from research to industrial applications, an accelerated development
and a further diversification of application areas is expected. AR paradigm
is fully exploited by including user-feedback. Actions or instructions
issued by the computer cause the user to perform actions changing the real
world, from which the computer may generate new augmentations [Klinker
et. al: 99]. In order to demonstrate interaction, different prototypes
have been made [Maes et. al.: 95], [Rose et. al: 95], [Behringer et. al.:
99], nevertheless a natural an intuitive way of user-input is still missing.
Currently, the most difficult problem in AR is the registration issue [Azuma-1:
97]. In contrast to virtual environments, visual AR applications require
a much higher registration precision. Indeed, the human eye is very sensitive
to a mismatch between virtual and real objects [Azuma-2: 97]. Existing
tracking devices are limited, either in their range capability (i.e. mechanical
or ultrasound tracker), or because of interference with metallic objects
(i.e. magnetic tracker).
Moreover,
reliable 3D-motion estimation from image features is an unresolved problem
in computer vision. One of the key technological challenges for creating
an augmented reality system is to maintain accurate registration and tracking
between real and computer-generated objects. Mobility is another requirement
for AR-applications. This problem is difficult because existing off-the
shelf wearable computer do not have an adequate interface support with
progressive scan cameras, which provide higher registration accuracy than
interlaced cameras.
Existing
solutions in documentation, maintenance, assistance to training
Paper
or microfilm documentation has long been the only solution to technical
manuals for complex mechanical systems. In the last years new systems (IETM),
based on CDROMS and PC’s, have replaced paper and microfilms. However,
IETM approaches have rather consisted in translating paper presentation
on digital media than providing paradigm shifts in the way documentation
is used. For training, virtual reality (VR) has been massively used in
all domains of industry, including mechanics. However, VR systems often
lack of realism in particular as far as kinesthesic interaction is concerned.
Moreover, in domains like those applications targeted by the project, the
necessary realistic interaction with objects to manipulate would be very
hard to attain. In the field of computer-guided maintenance based on AR,
different applications already exist (see above: repair of copiers, electric
wire assembly in aeroplanes, insertion of lock into a car door). Still,
most of these applications lack in regard to at least one of the following
properties:
-
registration accuracy,
-
two-way human computer
interaction,
-
mobility,
-
comprehensive multimedia
annotations support.
Innovation
in StarMate Project
StarMate
will develop a product bringing a paradigm shift in the way digital documentation
is used for assistance to maintenance. Indeed, making the user, the product
to maintain and its documentation cohabit and interact in the same working
environment is a first source of innovation. A second source of innovation
will be in the way the product provides information to the user. Instead
of browsing through data, the user will have a natural human-like dialogue
with its documentation system. He will query the system by speech and will
be guided through procedures to apply. He will have the possibility to
go backwards, ask for more information…
A
third innovation resides in the dual functionality of the system. Indeed,
the worker will use the system to perform maintenance, and he will also
have the possibility to train at his own pace on the same system. A fourth
innovation relies on the fact that the system will be specified, tested
and validate by end-users. The system will be developed to resolve operational
problems arising in industry. A fifth source of innovation is in the development,
combination and integration of leading edge technologies that will give
life to the product. Technical innovation of the system will reside in
the following aspects.
-
Resort to combined
use of active inertial tracking and 3D computer vision to alleviate current
limitations of registration.
-
Use of natural
human-like interaction to achieve user friendliness and preserve free hands
for working.
-
Use of visual and
audio augmentation, use of several multimedia supports.
-
Equipment preserving
user’s mobility.
-
Use of see-through
visualisation.
Beyond
innovative aspects of the project, the system will make use of off the
shelf material where possible. This will make the product affordable, will
allow reusability and interoperability, and will ease system upgradability.
|
