Hi Folks  
 
I'd like to put an idea across.  
 
I'm very new to motion tracking and I've noticed that certain shots are very hard to track. For example, you may have to deal with motion blur because the DOP likes low shutter speeds. Or maybe there are too many actors walking through the picture. Well... or you just have insufficient markers to track.  
 
But why not mount an additional HDV camcorder onto the camera rig, rotated 90°? With it, you could easily record an actor-free and unblurry high-shutter-take of the same shot. And, in addition, you can put some extra tracker points into the picture which you don't have to rotoscope in postproduction.  
 
After tracking the footage recorded by the HDV, isn't there a way to use this additional info to better solve the original footage? So that you could track even the hardest shots with just a few mouseclicks?  
 
Any thoughts about this?  
 
Greetings,  
j.

Hi,
This is definitely a usable workflow.  
If the 2 cameras are fixed together, they'll have the same motion
The only things that need to be done is then find the relative orientation, and add this to the solution. Scripts can be used to do that.
Bye

Hi Ronald
 
Thank you for your quick answer! May I ask you how that relative orientation can be added to the solution? Could you give me a hint? I think this is probably interesting for others, too.
 
Thanks a lot,
j.

You can do that with a script within MatchMover.
Once you know the relative orientation (like a delta rotation vector), you can simply add this delta to all the computed values before exporting.
Scripts are easy to use to do that. They have access to all the computed data and can also modify it. So it's easy !

Well, sounds logical. But I think there are probably some problems, especially when filming on 35mm:
 
- After transforming the camera's orientation and rotation from HDV to filmcam I don't have any trackers placed in my 3d space. Normally, I take the computed trackers to align my 3d geometry in my virtual world (to avoid sliding). But since the tracking was only done with the HDV footage I don't have trackers on my real shot.
- I don't know the focal width of the film camera. Sure I can ask the DOP and calculate it. But aren't there any inaccuracies due to the film transport mechanism or poor calibrated lenses?
- And, film transportation is shaky, so there is a movement in the optical center.
 
->Maybe just transforming the camera's orientation and rotation isn't enough.
Is there a way to take the solved and transformed HDV path and refine it to match the film footage? Like telling MM it should treat the precomputed path like the data from a motion control system for example?
 
I hope this is more or less comprehensible ;-)
 
 
Thank you very much,
j.