Hello,
 
I'm a new user to MMP,and just spent the last week tracking a variety of green screen shots. The biggest frustration I seem to encounter is the defining of user coordinates and I seem to get into the following loop.I'm doing the tracks as supervised, perform the 1st solve (thanks to Dom at support for pointing out the importance of correct reference frames); then looking at the 3D construct, I can see that some points are wrong, in particular the ground tracks which I know all lie on a shared Y value. In order to set up point relations, I first have to define a coordinate system, but I can only define a coordinate system relative to the solved points which I can see from the orthogonal views don't lie in suitable positions to generate a sensible coordinate system. Hope that makes sense! I don't understand why the option to use the existing internal coordinates aren't an option to establish point relations or survey points. I really hope I'm missing something here, but discussions with support seem to finally agree, after first insisting that single axis point relations didn't require user defined coordinates. Any ideas? Am I misunderstanding something ?
 
Thanks in advance.

Dear Neil,
 
can you provide the footage and your *.mmf-file on a ftp for download (jpg@90%)?
If it's confidential, send me a link via PM (privat message)
I'll have a look.
 
Best regards
 
Tim
 
tim.mendler@movematchers.net

Hi Tim,
Thank you for your quick response.
I'm not in work until Monday am, but I'll sort out the sequence then. I have had the following idea. What if I extract the X coordinate from one track(with a Known Y of 0, though given a slightly erroneous Y value by the solve), the Z of another track (again with a known Y of 0) either side of a track which the system has defined as 0,0,0 and entered these as survey points. Would that allow a coordinate system to be self generated and therefore in a coherent relationship to the camera and background plate ? At least then I could define a whole bunch of known Y values. Or would I just be compounding errors ?

Hi Neil,
might do, but not sure. Difficult to juge, without having seen your file.
I look forward to look at it on monday...  
N-joy your WE...! :-)
Best regards
Tim

Hey Tim,
 
You too.  
 
Thanks for the communication. Be in touch Monday
 
Regards,
 
Neil

Dear Neil,
after analysing your footage, it seems to me that the real problem is the leak of parallax.
The camera movement is almost just a pan, so there is no or verry little information for MM about the depth of the scene.
If you solve with a  "Nodal Pan" constraint, all the tracks will be on a spherical plane with its center in the camera.
The 3D-representation I get is very simular to this.
In such a case, single "helper" images (stills or single frames from another shot) showing the stage from another perspectiv are very usefull.
Otherwise, you probably will need to set up your coordinate system in the CG-application.
Hope that helps.
Best regards
 
Tim