History
Since June, 1st 2003 I'm working for Fraunhofer Gesellschaft, Institut fuer Intellegente Analyse- und Informationssyteme, Virtual Engeering. The actual project is called "Virtualizing of Scultures - Bernini digital" (see projectwebsite: Virtualizing of Scultures - Bernini digital ).
Our cooperationpartners are some guys from the art department of University of Siegen and the Gallerie Borghese in Rom.
Bernini built up four main statues for Galleria Borghese which are called David, Eneas and Anchise, The Rape of Proserpina and Appollo and Daphne in the the early years of the 17th century. Over the centuries those statues were moved arround due to event
puroses by ideas of the currators that time.
The actual placements now show the esthetical ideas of the 19th century. Our part is to get 3D-scanned each of the four statues by a range scanner.
We use a frameset of Eyetronix, which consits of a digital Canon EOS D60 SLR camera, a flashlight for projecting a grid through an optic, two nearly 270° x 270° moveable laserpointers mounted in a frame with some electronical stuff for taking grid and textur images. On my Bernini-Project-side you will find images of that frameset.
How does the Scanner work?
The flash is mounted in a distance between 15 and 50 cm left of the camera. That builts up a triangular between the cam itself, the flash and the target on which the grid is projected. After the images are stored on harddisk we use some programs to extract the grid from tif-images and safe them in Eyetronics fileformat ss3d.Each grid is till now extracted by Eytronic's Snachter wich is normaly use by fx-people from film companmies. The faceshot in Matrix I are done with the same equipment.
After detection, the grids are cleaned in a program which is called
"medit"
medit is a tool, written by Roland Kuck and myself, in the need of better analysing and correction of the grids. Eyetronix provides Shape Snatcher Suite with their Scanner.
Snachter doesn't provide adequat tools to cut the grid, add, delete, move and give coordinates to points in an easy way. They say, the mesh editing mode was just experimental and for internal use only. medit provides further functinality like recoordinating whole grids, or just parts of them, in space, because if grids, or just parts, are missplaced they get larger or smaller. Consequence of that is, they won't fit together.
If at least two grid patches exist as cleaned, or better "medited" patches, they have to be combined to one whole object.
Matcher, the tool which Eyetronics provides, can handle 52 patches. Our David consits of 452 patches now, so in the beginning we realized that we won't be able to load the whole statue into Matcher. Therefor we found other programs that can do that job.
One of it names Scanalys by Stanford University. Here they use "ply" as fileformat so we had to convert our ss3d to ply. The converter was written by Roland. With Scanalysis we are able to combine the patches to one whole statue. But this program has limitations to our task, too. It uses ICP for aligning but it can only align 2 patches at the same time. One solutions is putting 2 patches together and grouping it. Next step is taking the next patch aligning it to the group, ungroup it and regroup everything, then proceed this with the next patch. Every patch has its difficulties so in the end there may be errors of 15 to 20 cm. These errors come into existence because each pair has erros in the alignment in size of tenth of millimeter to millimeters. If you sum that up to 40 or much more patches you'll get that dimension of errors.
Now we have the whole statue consisting of single patches. For further use it is neccassary to have one patch. Therefor the Stanford-tool collection provides vrip. Its a program to combine the set of patches into one volume. Here resolution in milimeters is the main interessting value. The combining process takes time depending on the number of patches which belong to the model. About 500 patches take on a dual Xeon
processor with 2,4Ghz in an resolution of 0.2 dm between half an hour and 45 minutes
depending on the parameter used in vrip.
All in all, range scanning is one of the fastes ways to produce hires-3d-models which are most identical to the original.
This way of getting things digitized has limitations, too. One of them is, you are not able to scan matt dark objects or human hair.
Matt dark, nearly black objects is most probelematic because the grid which is projected has dark lines. If one woul'd use a inverted grid which is difficult to creat, it will do. Because you only can scan surfaces, hair can't be; it's not a surface it's loose. For light blond or white hair, there is a chance but don't calculate on it!
Scanalyse will be replace in the near future, too, because of its limitations to two-patch/group-registration.
A better way to deal with, is a solution where the patches are stored in a hirachy.
Scanalyse's "global registration" is better called, highly experimental and doesn't produce any good results.
My part in this project is the recronstruction of the statues, the geometry and writting programs to do that.
cu so long...
Orgun
07/03/2004
Due to some changes in
in the goals I'd like to reach, astronomy has to wait now for some
time. It has to wait till there will be money for further research especially equipment.
The brake won't be lost time, because I'm going to build up a radio
controlled helicopter, which will fly by
its own, based on a NEURAL NETWORK.
cu so long...
Our cooperationpartners are some guys from the art department of University of Siegen and the Gallerie Borghese in Rom.
Bernini built up four main statues for Galleria Borghese which are called David, Eneas and Anchise, The Rape of Proserpina and Appollo and Daphne in the the early years of the 17th century. Over the centuries those statues were moved arround due to event
puroses by ideas of the currators that time.
The actual placements now show the esthetical ideas of the 19th century. Our part is to get 3D-scanned each of the four statues by a range scanner.
We use a frameset of Eyetronix, which consits of a digital Canon EOS D60 SLR camera, a flashlight for projecting a grid through an optic, two nearly 270° x 270° moveable laserpointers mounted in a frame with some electronical stuff for taking grid and textur images. On my Bernini-Project-side you will find images of that frameset.
How does the Scanner work?
The flash is mounted in a distance between 15 and 50 cm left of the camera. That builts up a triangular between the cam itself, the flash and the target on which the grid is projected. After the images are stored on harddisk we use some programs to extract the grid from tif-images and safe them in Eyetronics fileformat ss3d.Each grid is till now extracted by Eytronic's Snachter wich is normaly use by fx-people from film companmies. The faceshot in Matrix I are done with the same equipment.
After detection, the grids are cleaned in a program which is called
"medit"
medit is a tool, written by Roland Kuck and myself, in the need of better analysing and correction of the grids. Eyetronix provides Shape Snatcher Suite with their Scanner.
Snachter doesn't provide adequat tools to cut the grid, add, delete, move and give coordinates to points in an easy way. They say, the mesh editing mode was just experimental and for internal use only. medit provides further functinality like recoordinating whole grids, or just parts of them, in space, because if grids, or just parts, are missplaced they get larger or smaller. Consequence of that is, they won't fit together.
If at least two grid patches exist as cleaned, or better "medited" patches, they have to be combined to one whole object.
Matcher, the tool which Eyetronics provides, can handle 52 patches. Our David consits of 452 patches now, so in the beginning we realized that we won't be able to load the whole statue into Matcher. Therefor we found other programs that can do that job.
One of it names Scanalys by Stanford University. Here they use "ply" as fileformat so we had to convert our ss3d to ply. The converter was written by Roland. With Scanalysis we are able to combine the patches to one whole statue. But this program has limitations to our task, too. It uses ICP for aligning but it can only align 2 patches at the same time. One solutions is putting 2 patches together and grouping it. Next step is taking the next patch aligning it to the group, ungroup it and regroup everything, then proceed this with the next patch. Every patch has its difficulties so in the end there may be errors of 15 to 20 cm. These errors come into existence because each pair has erros in the alignment in size of tenth of millimeter to millimeters. If you sum that up to 40 or much more patches you'll get that dimension of errors.
Now we have the whole statue consisting of single patches. For further use it is neccassary to have one patch. Therefor the Stanford-tool collection provides vrip. Its a program to combine the set of patches into one volume. Here resolution in milimeters is the main interessting value. The combining process takes time depending on the number of patches which belong to the model. About 500 patches take on a dual Xeon
processor with 2,4Ghz in an resolution of 0.2 dm between half an hour and 45 minutes
depending on the parameter used in vrip.
All in all, range scanning is one of the fastes ways to produce hires-3d-models which are most identical to the original.
This way of getting things digitized has limitations, too. One of them is, you are not able to scan matt dark objects or human hair.
Matt dark, nearly black objects is most probelematic because the grid which is projected has dark lines. If one woul'd use a inverted grid which is difficult to creat, it will do. Because you only can scan surfaces, hair can't be; it's not a surface it's loose. For light blond or white hair, there is a chance but don't calculate on it!
Scanalyse will be replace in the near future, too, because of its limitations to two-patch/group-registration.
A better way to deal with, is a solution where the patches are stored in a hirachy.
Scanalyse's "global registration" is better called, highly experimental and doesn't produce any good results.
My part in this project is the recronstruction of the statues, the geometry and writting programs to do that.
cu so long...
Orgun
07/03/2004
Due to some changes in
in the goals I'd like to reach, astronomy has to wait now for some
time. It has to wait till there will be money for further research especially equipment.
The brake won't be lost time, because I'm going to build up a radio
controlled helicopter, which will fly by
its own, based on a NEURAL NETWORK.
cu so long...
Orgun
05/29/2001:
This websites are about my work and special interest. They contain
information about galaxyclusters (my way to find them) and reduction of
CCD-frames.
I'm working at Max-Planck-Institute fuer Extraterretrische Physik,
Garching since July 2000. Also i'm a student of Ludwig-Maximilians
Universitaet at Munich, where i studied physics vor two years now.
In the early days here I learnt how to reduce optical ccd-frames by
usin IRAF.
Basic technics and basic knowlegde like: the ccd-ensemble consists of
what? what are the connexion between BIAS, FLAT and DATA framse? have
been the preparation of the following reduction of images takes at Skinakas-Obersavtory
on Crete.
The project I'm still working on is called 'Analysis of optical sky
accommodations of galaxy cluster (Rosat ALL Sky Survey identification)'.
My function is, creating a computer-assited knowledge pipeline, in
order to be able to detect galaxy clusters in crowded star fields.
Reduction methods are just the first parts, step one, on my way to
hell... :-)
After having reduced a total oberservation run of Skinakas, they sent
me to that observatory to get some images myself, cause I got to know
much about the problems of their chip. In order to take several images
in R-band we had just two nights of a hole week due bad wether.
While I was reducing the first run, using those early evening-periods
of non-concentration for chatting with people I got to know a nice
girl, who now shares a relationship with me.
'God bless the internet'... (well done Eugen ;-))
Back to duty now! After havin reduced that second run and a third of a
colleague, who took some images at La Silla's Danish 1.5 m
Eso-Teleskop, I started phase two of our journey towards a
fullautomatic reduction and analizing routine for optical data.
I was introduced to a source extracting, catallog producing programm
called 'Sextractor' (written by E. Bertin).
For making statisics about number density distributions the first part
of this stepp is writing programms to understand features given by
sextractor. After figuring out the quality of extraction, next step is
counting the seperated galaxies in rings around a selected position on
the chip. Last part then will be
comparing ccds of different galactical longitudes.
Phase three can be title with: 'Final automation of reducing and
analyses process'...
By reducing ccd-images of various telescops I remember seeing many
strange artifacts with I'm going to discuss on my page two. There will
be a link to a ccd-error-collekting page,too. For not having to quote
the content of this page I like to refer the readers to that page,
(it's a good one :-)!
That's all for now, folks!
cu
Orgun