RE: reverse engineering benchmarking

From: Anshuman Razdan (razdan@asu.edu)
Date: Wed Aug 28 2002 - 07:52:05 EEST


Dear Bernard
        Thanks for the note. I agree with you. So what you are saying is
there should be a different benchmarking for scanners (surface scanners
such as point, laser, etc. and volume scanners such as CT, MRI etc.) and
different benchmarking for surface fitting. I am not sure about the
accuracy of CT/MRI scanners in the sense that they collect lots of data
but then output filtered data as 2D slices from which software like
materialize create the 3D surfaces. The problem with "live" subject is
-- how the hell are you going to measure internal organs for comparison?
:-). I am very much for benchmarking of surface scanners. In fact having
bought four scanners for my lab so far I have experienced the difficulty
of getting straight answers about resolution, error tolerance etc. If my
lab can help I will be happy to participate. We have scanned over 800
bones from different museums, over 100 cermaic vessels from Hopi (around
1000 AD) era, etc. and looking for diverse kinds of data. We have two
cyberware (M15 and 3030) and two portable LDI scanners.

        As far as surface fitting software I am not sure there will ever
be a straight forward method for benchmarking since there is lot of user
interaction involved.

        I was not able to attend your conference this year but perhaps
can participate next year.

Best.

AR

> -----Original Message-----
> From: Alain BERNARD [mailto:Alain.Bernard@irccyn.ec-nantes.fr]
> Sent: Tuesday, August 27, 2002 10:31 AM
> To: razdan@asu.edu
> Cc: rp-ml@rapid.lpt.fi
> Subject: RE: reverse engineering benchmarking
>
>
> Dear Anshuman,
>
> I appreciated your comment and input.
>
> What I had in mind was more related to 3D scanning than the
> complete chain. I agree with you about the influence of the
> methodology of use of the models and tools. Most often the
> quality of the result depends on the human decision related
> to the process (the strategy) of scanning. The calibration
> phase is also very deterministic for the quality of the result.
>
> How to compare the result, this is a good question because
> the interpretation of the results is very often made using
> the software you buy with the sensor. Due to the wide range
> of applications (very small objects to very large ones, one
> view or multiple views, external and/or internal shapes,
> etc...) it seems difficult to define generic standard part
> and procedure...
>
> In my opinion, the problem exists also for accuracy, geometry
> and dimension parameters regarding layered manufacturing. The
> result depends very often on the way you choose for the
> position of the part, the process parameters and the material.
>
> Concerning RE, one can refer to the interesting 6 pages
> synthesis you wrote in Terry Wohlers report. There are also
> two appendixes presenting lists of sensors and softwares.
>
> As people from Numerisation 3D replied before this input, one
> has also to consider Scanning conference in Paris and all the
> different thematic aspects related to it as a very
> interesting reference, every year. The papers are mainly
> proposed by academics and users and a place in the program
> exists for debate with vendors of sensors and softwares.
>
> Well, this is my feeling.
>
> Best Regards,
>
> Prof. A. BERNARD
>
> At 09:21 27/08/2002 -0700, you wrote:
> >It is very hard to come up with bench marking for RE. I will
> give you a
> >few reasons. First lets out line the steps usually performed for RE.
> >
> >A) Digitizing: Starting with a control part it must be
> digitized. Right
> >away you are embroiled in controversy sine very few
> digitizing vendors
> >really will give you error measures. So now you have built
> an Epsilon
> >error in to the digitzed model. Further, with digitizers
> that align and
> >merge views, errors creep in when resolving those points along the
> >seams. That is usually around sharp edges or feature lines. The
> >digitizing post process also involves cleaning data, editing support
> >out of the data. Error is now Epsilon+
> >
> >B) Point cloud to Triangle meshes. Again done as a post
> process or in
> >another application this is usually the first step. This further
> >involves despiking, refining, removing outliers, other noise removal
> >filters. Error is now Epsilon++.
> >
> >C) The last (major) step being putting Bspline/NURB surfaces on the
> >mesh. Most software (no offense) work on simple planar
> projection for
> >creating parametrization, depend on user to give the number
> of control
> >points in U, V direction. The error calculation which can be very
> >cumbersome is also not done accurately (although it suffices in most
> >cases but in bench marking one create cases that will fail
> miserably).
> >This is the major problem in adding to the erorr which adds
> couple of
> >++ to Epsilon so its now Epsilon++++.
> >
> >D) Not included in this process is the floating point
> numerical error
> >which creeps in all the processes above shaving a few
> precision digits.
> >
> >E) The final step -- how would you check and compare. The inspection
> >software component resamples the Nurb/Bspline surfaces at
> sufficiently
> >large sample and compares agains the triangle mesh oin B or
> the point
> >cloud against the tessallated surface from C (mind you not the
> >continuous surface that is the NURBS model since it would be
> very very
> >expensive).
> >
> >My point (I donít want to depress any one here) is that
> unlike RP where
> >surface finish can be measured with high degree of
> precision, RE is a
> >process that is frought with lots of bumps. Its not the fault of
> >digitizers or software vendors -- the process has several
> variabilities
> >that are not repeatable. For example -- laying the NURBS grid on the
> >traingle meshes. Two people will do it differently and will come out
> >differently. Except for simple parts the RE softwares do not
> do a good
> >job of laying the grid -- it always requires tweaking, adding or
> >removing feature lines/edges, etc.
> >
> >So in conclusion -- the digitizing and software is getting
> better but
> >benchmarking I am not sure would give right answers. The problem is
> >that it can be more misleading than not having benchmarks.
> Does any one
> >really get 10 to the minus 5 precision in the final part ?
> >
> >However, this is a good debate and I am happy people are thinking in
> >this direction.
> >
> >Best
> >
> >AR
> >
> >Dr. Anshuman Razdan
> >Director, PRISM
> >http://3dk.asu.edu
> >Phone: (480) 965 0483 (Tina)
> >Office: GWC 574
> >Arizona State University
> >Tempe AZ 85287-5906
> >
> >
> >> -----Original Message-----
> >> From: owner-rp-ml@rapid.lpt.fi
> >> [mailto:owner-rp-ml@rapid.lpt.fi] On Behalf Of Ping Fu
> >> Sent: Tuesday, August 27, 2002 7:22 AM
> >> To: 'Yasser Hosni'; Alain.Bernard@irccyn.ec-nantes.fr;
> >> georgeh@slingshotpdg.com
> >> Cc: rp-ml@rapid.lpt.fi
> >> Subject: RE: reverse engineering benchmarking
> >>
> >>
> >> There is a need for an annual report on RE like what Terry
> >> does for RP. We need category groups, then within a
> >> sub-group, BM can be done meaningfully.
> >>
> >> Ping
> >>
> >> -----Original Message-----
> >> From: owner-rp-ml@rapid.lpt.fi
> >> [mailto:owner-rp-ml@rapid.lpt.fi] On Behalf Of Yasser Hosni
> >> Sent: Monday, August 26, 2002 8:23 PM
> >> To: pfu@geomagic.com; Alain.Bernard@irccyn.ec-nantes.fr;
> >> georgeh@slingshotpdg.com
> >> Cc: rp-ml@rapid.lpt.fi
> >> Subject: RE: reverse engineering benchmarking
> >>
> >>
> >> Hi:
> >> Reverse Engineering (RE) is a broad field. For example,
> >> surface scanning is considered RE, so as the CT/MRI. In
> >> conducting bench marking (BM) studies, one need to consider
> >> categorizing RE so that a meaningful BM can be developed.
> >> In addition, the technologies employed are different, with
> >> advantages and deficiencies in certain areas. Hence there is
> >> a need for BMing within each technology. Examples include
> >> regular laser scanning (for capturing the geometrical
> >> configuration/ measurements) and color laser scanning (for
> >> capturing the colors objects). I am sure BM can be developed,
> >> however, one has to be careful in its development. Take care. YH
> >> Yasser Hosni, Ph.D., PE.
> >> Martin Marietta Distinguished Professor
> >> Industrial Engineering and Management Systems
> >> University of Central Florida
> >> 4000 Central Florida Blvd.
> >> Orlando, FL 32816
> >> Tel (407) 823-5817
> >> Fax (407) 823-3413
> >> E-mail: yhosni@mail.ucf.edu
> >>
> >> >>> "Ping Fu" <pfu@geomagic.com> 08/26/02 01:46PM >>>
> >> In my extensive market research, I have not seen complete
> >> benchmark being done in this industry sector. If anyone is
> >> interested in doing so, it surely will benefit the community.
> >> I've seen a few benchmark done by university students,
> >> customers, but all are incomplete and some are not allowed to
> >> be distributed.
> >>
> >> Ping
> >>
> >> -----Original Message-----
> >> From: owner-rp-ml@rapid.lpt.fi
> >> [mailto:owner-rp-ml@rapid.lpt.fi] On Behalf Of Alain BERNARD
> >> Sent: Monday, August 26, 2002 1:12 PM
> >> To: George Hatzilias
> >> Cc: RPML
> >> Subject: RE: reverse engineering benchmarking
> >>
> >>
> >> Dear Giorgos ,
> >>
> >> We might be impressed by the content of such websites and of
> >> the numerous of sensors. But I wonder if a general method of
> >> benchmarking has been proposed for all these sensors and
> >> technology. I also know some initiatives but for some
> given sectors.
> >>
> >> Any information about that?
> >>
> >> Best Regards,
> >>
> >> Prof. A. BERNARD
> >>
> >> At 10:37 21/08/2002 -0400, you wrote:
> >> >Simon,
> >> >
> >> >These lists are a good start:
> >> >www.simple3d.com http://paraform.com/template.asp?pageid=2.6.3
> >> >http://www.geomagic.com/support/resources/3scanners.pdf
> >> >
> >> >Georgia Tech has done some work on standardized benchmarking
> >> of these
> >> >systems, contact Dr. Tom Kurfess, a professor in metrology
> >> and head of
> >> >the precision manufacturing research lab at
> >> tom.kurfess@me.gatech.edu.
> >> >I beleive he has published a few papers on benchmarking and
> >> comparing
> >> >such systems.
> >> >
> >> >I would also be glad to share from my experience as a former
> >> engineer
> >> >from Paraform, where I was exposed to most every major scanner
> >> >available at some point or other. I now help operate a
> >> scanning service
> >>
> >> >beareau scanning from small micron scale parts to full size
> >> vehicles,
> >> >glad to share any insight on particular scanners.
> >> >
> >> >Thanks,
> >> >Giorgos Hatzilias
> >> >Slingshot PDG
> >> >3675 Crestwood Pkwy Suite 400
> >> >Duluth, GA 30096
> >> >404-317-9312
> >> >www.slingshotpdg.com
> >> >georgeh@slingshotpdg.com
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >----- Original Message -----
> >> >From: "simon ram" <>
> >> >To: <rp-ml@rapid.lpt.fi>
> >> >Sent: Sunday, August 18, 2002 5:37 AM
> >> >Subject: reverse engineering benchmarking
> >> >
> >> >
> >> >>
> >> >> Does anybody know or has employed a benchmark to evaluate the
> >> >> performances
> >> >of the different reverse engineering systems ?
> >> >>
> >> >> I would be very interested in receiving this information
> >> and if is it
> >> >possible the model file or information about it.
> >> >>
> >> >> I'm interested mainly interested in working with
> structured light,
> >> >> laser
> >> >scanner and touch probe systems.
> >> >>
> >> >> Thank you very much for the help, any suggestions or information
> >> >> about
> >> >reverse engineering benchmarking will be very interesting
> >> >>
> >> >> Regards
> >> >>
> >> >> Simon
> >> >>
> >> >>
> >> >>
> >> >>
> >> >>
> >> >> ---------------------------------
> >> >> Do You Yahoo!?
> >> >> HotJobs, a Yahoo! service - Search Thousands of New Jobs
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >For more information about the rp-ml, see
> http://rapid.lpt.fi/rp-ml/
> >> >
> >>
> -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> >> Prof. Alain BERNARD
> >> IRCCyN UMR CNRS 6597
> >> Ecole Centrale de Nantes
> >> 1, rue de la NoŽ BP 92101
> >> 44321 - NANTES Cedex 03 (France)
> >> tel : + 33 (0) 2 40 37 69 53
> >> fax : + 33 (0) 2 40 37 69 30 -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> >>
> >> For more information about the rp-ml, see
> http://rapid.lpt.fi/rp-ml/
> >>
> >>
> >> For more information
> about the rp-ml, see http://rapid.lpt.fi/rp-ml/
> >>
> >> For more information about the rp-ml, see
> http://rapid.lpt.fi/rp-ml/
> >>
> >>
> >> For more information
> about the rp-ml, see http://rapid.lpt.fi/rp-ml/
> >>
> >
> >
> -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> Prof. Alain BERNARD
> IRCCyN UMR CNRS 6597
> Ecole Centrale de Nantes
> 1, rue de la NoŽ BP 92101
> 44321 - NANTES Cedex 03 (France)
> tel : + 33 (0) 2 40 37 69 53
> fax : + 33 (0) 2 40 37 69 30 -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>

For more information about the rp-ml, see http://rapid.lpt.fi/rp-ml/



This archive was generated by hypermail 2.1.4 : Tue Jan 21 2003 - 20:14:12 EET