J. B. Thomine, TAKASHI KOEZUKA, KENJI AKIYAMA, YUKIO KOBAYASHI
INTERNSHIP REPORT
Abstract:During my internship in ATR, I was with the Communication
Systems Laboratories, in the Artificial Intelligence Department.
One of the goals of this department is the automatic acquisition
and reconstruction of three dimensional objects (further use of
such techniques is video-shopping and so on).
Two projects are now under study: one uses the slit-ray
projection method and the other one is based on Moire topography. I was included in the team working on the second subject: one researcher, Mr Koezuka, two programmers, Mr Ohuchi and
Mr ura, and myself.
Moire topography permits to visualise the equi depth lines
of an object, therefore its usefulness for obtention of 3D data is
obvious. It is quite an old technique (back to the late sixties)
but, before our research, the process leading to the obtention of
fringes had always been completely optical, and the result a picture
useful for human inspection, but inadequate for the formation
of a computer database (because of noise, change of contrast and
unability to distinguish convex from concave parts). Therefore, one of the goals of the project was to define a digital process
in an image processing system leading to the obtention of fringes
easy to acquire by the high-speed computer establishi
ng the
database.
I was given the task to draw the outlines of this process and
to implement it on the VICOM image processing system. I have been
working on this subject for the eight months of of my internship
and I could define several solutions to the main problem (generation of the fringes). One of the algorithms is now
in use for the acquisition of 3D data. I have also found a very
simple method to distinguish between convex and concave parts, which lead to the implementation of a thinning algorithm
specific to the Moire process and with very good precision characteristics.
With those results, we have registered a Japanese patent, and also written a paper for a Japanese proceeding and a prepaper for an international conference on "Robotics and Automation" to be held next year.
At the end of my internship, when we could dispose of a
precise optical stage, we realized we had to face some new problems due to inherent errors in the video acquisition system(distortion). These new problems, although I have some ideas they may be drastically reduced by the use of another video input system, still remain to be clearly solved as they affect in an annoying way the process of reconstruction.