Big frustration here.
 
I placed my 2D markers on this one footage.
Solve for camera and now want to create a constraint on a few points that share the same z coordinate (points are on a wall).
 
SO i create the new constraint.
Select the tracks that are part of the constraint, select share Z, select unknown, and enable.
 
My constraint is now created. Now from what i understand, i still need to select frames and then tell MMP that the newly created constraint is to be valid on the selected tracks only for those selected frames.
 
Well i want this constraint to be always valid so i marquee (click and drag rectangle) in the timeline (or in the tracks window) the entire shot. The entre timeline becomes yellow.
 
Now if i go to 3D tracking > Edit Constraint, the three options (add frames, remove frames, toggle) are greyed out.
 
I did go check in the camera window that the time range was set from start of footage to end of footage (because i selected everything).
I also made sure that 'use selection' was selected. Then it wasn't working so i also tried with clicking on global time.
 
No matter what i do i cannot find a way to add the frames to my constraint. What tricky manipulation am i missing? Read the manual multiple times, i'm running out of ideas....
 
What am i doing wrong? (using MMP4

Hi,
 
You only have to select frames for camera constraints.
 
In your case, this is a "point relation" and it is valid for the whole sequence: your 3d tracks are static, their coordinates don't change, no need to select any frames.
 
Regards,
 
Niko

ok thanks a lot.
 
That's what the manual says after all. Selecting frames and adding them to a constraint is only mentionned in the section that deals with camera constraints.
 
But i was confused because i am on the impression that my wall constraint is not being taken into account during the solving process. Its just that after creating my plane (=wall) constraint and solve for the camera, the result seems to be exactly the same as the one i get when not defining the constraint. It is as if definining and using this constraint had no effects.
 
All of these points that are supposed to be on the same plane are still all over the place in the 3D solution.
 
I thought that defining a constraint would force MMP to find a solution with those points being on the same plane.
 
The set has been image modeled with image modeler and in image modeler i can create faces between those points and then import the geometry in MMP. Then use the rubber band to associate each vertex of the geometry to the 2D/3D tracks. I as going to try this option but will that lead me to a better solution?
 
Wouldn't that be the same as defining the planar constraint? Defining a planar constraint or importing a plane, isn't that the same idea. They both basically tell MMP that those points are non the same plane.
 
But in the end how do i force those points to be on the same vertical plane?
 
 
 
For your information: My shot does not have a lot of parallax change. A push-in dolly with only 4 markers not part of the background wall (4 stands, 1 marker per stand). However many markers on the wall.
No idea what the camera focal length was eihter. (constatn though) Solving with various initial focal length give the same reaprtition of markers in 3 space (different spacing but spatial repartition is the same).
 
Thanks for your help.

just giving a bump tp this post.
 
My 2D tracks are great, residual less than 0.2 for all tracks.
 
Solution in 3D however doesn't reflect reality.
 
Even if i define this plane constraint, 3D locators are still not aligned on a single plane.
 
So should i conclude that constraints on points are not very powerful tools for helping the camera solve? just a way to fine-tune your results assuming your solution is already quite good?

Hi,
 

 
Yep, you're right. If it's not you have to define a coordinate system for the relations to be used during camera solving.
 
If your points are not on the same plane in the solution, it means ether that the relation is disabled, a coordinate system has not been set or the solution has not been re-computed.
 
The suggested workflow is the following one:
- tracking
- solving
- define a coordinate system
- define relations if you wished
- solving
 

 
Yes it is.
 

 
This can happen if your shot has not enough parallax (a nodal pan for example). Try to add helpers frames if you have some or use survey points or relations to help MatchMover compute good 3d: i know this is what you are trying to do but i hope it will work once you will have defined a coordinate system.
 
Hope it helps,
 
regards,
 
Niko

i found out what was confusing me.
 
If I just define a coordinate system, the planar constraint (points on wall), then solve, the solution is good.
 
Looking in 3D and rotating, the points are indeed part of the same plane.
It all looks good until i do 'map world on camera' which i sued to do automatically to look in 3D.
 
After doing the map world on camera, those points are not on the same plane anymore.... i guess i don't understand what map world on camera  does. I though it was just a way to transfer a coordinate system to the camera.
 
The solution before doing map world on camera is good (correct repartition of the markers in 3D space) but after doing 'map world on camera', they do not respect hte planar constraint anymore.
 
Note: my planar constraint has more than 3 points (3 points always share a common plane). So it is not like hte plane is jsut inclined realtive to the new coord system. The 10 points are all in random places.