< Previous | Contents | Manuals Home | Boris FX | Next >
Alignment Versus Constraints
With a small, well-chosen, set of constraints, there will be no conflict among them: they can all be satisfied, no matter the details of the point coordinates. This is the case for the 3-tracker recommended method.
However, this is not necessarily the case: you could assign two different points to both be the origin. Depending on their relative positions, this may be fine, or a mistake.
SynthEyes has two main ways to approach such conflicts: treating the coordinate system constraints as suggestions, or as requirements, as controlled by the Constrain checkbox on the Solver control panel.
For a more useful example, consider a collection of trackers for features on the set floor. You can apply constraints telling SynthEyes that the trackers should be on the floor, but some may be paint spots, and some may be pieces of trash a small distance above the floor.
With the Constrain box off, SynthEyes solves the scene, ignoring the constraints, then applies them at the end, only by spinning, rotating, and scaling the scene. In the example of trackers on a floor, the trackers are brought onto an average floor plane, without affecting their relative positions. The model is fundamentally not changed by the constraints.
On the other hand, with the Constrain checkbox on, the constraints are applied to each individual tracker during the solving process. Applied to trackers on a floor, the vertical coordinate will be driven towards zero for each and every such tracker, possibly causing internal conflict within the solving process.
If you have tracked 3 shadows on the floor, and the center of one tennis ball sitting on the floor, you have a problem. The shadows really are on the floor, but the ball is above it. If all four height values are crunched towards zero, they will be in conflict with the image-based tracking data, which will be attempting to place the tennis ball above the shadows.
You can add poorly chosen locks, or so many locks, that solving becomes slower, due to additional iterations required, and may even make solving impossible, especially with lens distortion or poor tracking. By definition, there will always be larger apparent errors as you add more locks, because you are telling SynthEyes that a tracker is in the wrong place. Not only are the tracker positions affected, but the camera path and field of view are affected, trying to satisfy the constraints. So don’t add locks unless they are really necessary.
Generally, it will be safer to leave the Constrain checkbox off, so that solving is not compromised by incorrectly configured constraints. You will want to turn the checkbox on when using multiple-shot setups with the Indirectly solving method, or if you are working from extensive on-set measurements. It must be on to match a single frame.
©2024 Boris FX, Inc. — UNOFFICIAL — Converted from original PDF.