Photogrammetry software needs to make assumptions when it has nothing more than 
color information.

Specular highlights are going to be clipped/clamped to white or something 
similar.  Therefore, if the software sees white, how is it to know whether it 
is a specular highlight vs. just something colored white?  That is the problem. 
 In the case of transparency, how does it know the difference between a textile 
pattern vs. colors poking through multiple surfaces?

Providing multiple angles of the same subject can help resolve those issues 
because the software can cross reference the details as perspective changes and 
see the parallax shift, but if the subject is highly reflective or glossy the 
specular highlight will travel around the surface as you change angles.  
Therefore, you may need to feed additional images to the software to give it 
more information to resolve the problems, and those problems may not be fully 
resolvable unless you make adjustments to lighting - or do what a lot of the 3D 
scanner companies do in their demos - apply dust/powder to the subject to 
remove glossiness and transparency.  At that point It's not much different than 
making multiple passes with a 3D scanner and using registration points to align 
the geometry after the fact.

These issues are less problematic with 3D scanners because the sensor has more 
information at its disposal from scanning the subject directly and can 
differentiate using alternate information such as intensity of light, by using 
different wavelengths such as infrared, or different technologies altogether 
such as sound waves.  Actual technique varies with the scanner.


Matt



From: [email protected] 
[mailto:[email protected]] On Behalf Of Manuel Huertas 
Marchena
Sent: Wednesday, June 04, 2014 11:20 AM
To: softimage list
Subject: RE: Photogrammetry - what do you use?

Hi Stephen,

I am curious to why it does not work with transparent/shiny objects (havent 
really done any test with those kind of surfaces..). Do you mean that
the calibration for the point cloud isn't accurate? what software & workflow 
are you using?


________________________________
Date: Wed, 4 Jun 2014 14:13:28 -0400
Subject: Re: Photogrammetry - what do you use?
From: [email protected]<mailto:[email protected]>
To: [email protected]<mailto:[email protected]>
Be aware that there is no Photogrammetry solution, that I have found that will 
deal well with
transparent and/or shiny objects. I do a lot of product modeling, from 
prototypes, and I have
not found any Photogrammetry solution that works better than taking front, 
side, top, and 3/4
view photos, and using the rotoscope function in my views. I have tried many.

I am hoping that the new 3D scanning for ipad will be better, but it looks 
similar to other methods.
https://www.kickstarter.com/projects/occipital/structure-sensor-capture-the-world-in-3d

I wish someone would come out with a 3D scanner that is based on sonar 
principles.
Image based 3D scanning has so many issues. The laser scanners are nice, but 
have
issues with undercuts as well as transparent and shiny surfaces.

If you do find a solution, that works well, please post back here.



On Wed, Jan 29, 2014 at 10:08 AM, Marc-Andre Carbonneau 
<[email protected]<mailto:[email protected]>> 
wrote:
Hello friends,

I am currently investigating photogrammetry and would love to get your advices, 
opinions, experiences with such systems.
What hardware do you use? Which software? Best practices?

Thank you for any info!
MAC


[cid:[email protected]]<https://mdc-web-tomcat17.ubisoft.org/confluence/display/technologygroup/Home+Passenger>

Marc-André Carbonneau
Product Specialist








--
Best Regards,
  Stephen P. Davidson
       (954) 552-7956
    [email protected]<mailto:[email protected]>
Any sufficiently advanced technology is indistinguishable from magic
                                                                             - 
Arthur C. Clarke
[http://www.3danimationmagic.com/3Danimation_magic_logo_sign.jpg]<http://www.3danimationmagic.com>

Reply via email to