Constraining Camera Position and Motion

Constraining Camera Position and Motion

Sometimes you may already know some or all of the path of the camera, for example,

·        it may be available from a motion controlled camera,

·        the camera motion may be mechanically constrained by a camera dolly,

·        you may have measured some on-set camera data to determine overall scene sizing, or

·        you may have already solved the camera path, then hand-edited it for cleanup.

SynthEyes lets you take advantage of this information to improve a solution or help set up the coordinate system, using the trajectory lock controls at the bottom of theSolver Control Panel, theHard and Soft Lock Controldialog, and the camera's seed path information.

Warning:using camera position, orientation, and field of view locks is avery advanced topic. You need to thoroughly understand SynthEyes and the coordinate system setup process, and have excellent mental visualization skills, before you are ready to consider camera locks. Under no circumstances should they be considered a way to compensate for inadequate basic tracking skills.

Concept and Terminology

SynthEyes allows you to create locks on path (X, Y, and/or Z translation), rotation (pan, tilt, and roll), and field of view. You can lock one or more channels, and locks are animated, so they might apply to an entire shot, a range of frames, or one frame.

Each lock forces the camera to, or towards, the camera's seed path. The seed path is what you see before solving, or after clearing the solution. You can see the seed path at any time using theView/Show seed pathmenu control, or the button on the Hard and Soft Lock Control dialog.

Locks may be hard or soft. Hard locks force the camera to the specified values exactly (except if Constrain is off), similar to pegged trackers. Soft locks force the camera towards the specified value, but with a strength determined by a weight value.

Locks are affected by theConstraincheckbox on the solver panel, similar to what happens with trackers. With Constrain off, locks are appliedaftersolving, anddo notwarp the solution. All soft locks are treated as hard locks. With Constrain on, locks are applied before and during solving, soft locks are treated as such, and locksdowarp the solution. Field of view locks are not affected by Constrain, and are always applied.

Camera position locks are more useful than orientation locks; we’ll consider position locks separately to start with.

You can also constrain objects, but this is even more complex.

Basic Operation

Set up generally proceeds as follows:

1.   If you have not already attempted to solve the scene, go to step 5.

2.   Go to the Solver Control Panel. Click themorebutton to bring up theHard and Soft Lock Controlpanel.

3.   Position and animate the camera as desired, creating a key on each frame where you want the position to be constrained. The Get buttons can help with this.

4.   Turn on the L/R, F/B, and/or U/D buttons as appropriate depending on the axes to be constrained — these stand for left/right, front/back, and up/down respectively.

5.   Adjust the Constrain checkbox as needed. The camera position constraints behave similarly to constraints on the trackers: if the Constrain checkbox on, they are enforced exactly during the solve, but if the Constrain checkbox is off, they are enforced only loosely after the completion of the solve. Loosely means that they are satisfied as best as can be, without modifying the trajectory or overall RMS error of the solution.

The result of this process is to make the camera match the X, Y, and/or Z coordinates of the seed path at each key. This basic setup can be used to accomplish a variety of effects, as described above and covered in more detail below. At the end of the section, we’ll show some even more exotic possibilities.

Using a Camera Height Measurement

Suppose the camera is sitting on a moving dolly in a studio, and you measured the camera (lens's) height above the floor, and you have some trackers that are (exactly) on that floor. You can use the height measurement to set up the scene size as follows:

1.   Show the seed path: View/Show Seed Path menu item

2.   At frame 0, position the camera at the desired height above the ground plane: 2 meters, 48 inches, whatever.

3.   Turn on the U/D button on frame 0,turn it back off at frame 1.

4.   Set up a main coordinate system using 3 or more trackers on the floor. Make sure tonotcreate a size constraint in the process: if using the *3 button on the Coordinate system panel or the Coord button on the Summary panel, select the 2^nd(on-axis) tracker, and in the Coordinate panel, change it from Lock Point (at 20,0,0) to On X Axis or On Y Axis.

5.   Solve with Go! on the Solver panel

Note that you can use whatever more complex setup you like in step 4, as long as it completely constrains both the translation and rotation, but not the size.

WARNING: You might be tempted to think “Hmmm, the camera is on a dolly, so the entire path must be exactly 43 inches off the floor, let me set that up!” (by not turning U/D back off). But this is almost alwaysa bad idea! The obvious problem is that the dolly track is never really completely flat and free of bumps. If the vertical field of view is 2 meters, and you are shooting 1080i/p HDTV, then roughly your track must beperfectly flat to 1 millimeteror so to have a sub-pixel impact. If your track is that flat, congratulations.

The conceptually more subtle, but bigger impact problem is this: a normal tripod head puts the camera lens very far from the center of rotation of the head—roughly 1 foot or 0.25 meter. As you tilt the head, the position of the camera increases and decreases up to that much in height! Unless your camera does not tilt during the shot, or you have an extra-special nodal-pan head, the camera height will change dramatically during the shot.

A Straight Dolly Track Setup

If your camera rides a straight dolly track, you can use the length of that track to set the scale, and almost the entire coordinates system if desired. While the camera height measurement setup discussed above is simpler, it is appropriate mainly for a studio environment with a flat floor. The dolly track setup here is useful when a dolly track is set up outdoors in an environment with no clearly-defined ground plane—in front of a hillside, say.

For this setup, you should measure the distance traveled by the camera head down the track, by a consistent point on the camera or tripod. For example, if you have a 20’ track, the camera might travel only 16’ or so because there will be a 2’ dead zone at each end due to the width of the tripod and dolly. Measure the starting/ending position of the right front wheel, say.

Next, clear any solved path (or click View/Show seed path), and animate the camera motion, for example moving from 0,0,0 at the beginning of the shot to 16,0,0 at the end (or wherever it reaches the maximum, if it comes back).

You now have two main options: A) mostly tracker-based coordinate setup, or B) mostly dolly-based coordinate setup, for side-of-hillside shots.

For setup A, turn on only the L/R camera axis constraint checkbox on the first and last frames (only). The X values you have set up for the camera have set up an X positioning for the scene, so when you set up constraints on the trackers, they should constrain rotation fully, plus the front/back and up/down directions—but not the L/R direction since that would duplicate and conflict with the camera constraint (unless you are careful and lucky).

For setup B, turn on L/R, F/B, and U/D on the first and last frames (only). You should take some more care in deciding exactly what coordinate values you want to use for each axis of the animated camera path, because those will be defining the coordinate system. [By setting keys only at the beginning and end of the shot, you largely avoid problems with the camera tilting up and down—at most it tilts the overall coordinate system from end to end, without causing conflicting constraints.]

If the track is not level from end to end, you can adjust the beginning or ending height coordinate of the tracker as appropriate. But usually we expect the track to have been leveled from end to end.

With X, Y, and Z coordinates keyed at the beginning and end of the shot, you have already completely constrained translation and scale, and have constrained 2 of the 3 rotation axes. The only remaining unconstrained rotation axis is a rotation around the dolly.

To constrain this remaining rotation requires only a single additional tracker, and only its height measurement! On the set, you should measure the relative height of a trackable feature compared to the track (usually this will be to the base of the track, so you should also measure the height of the camera versus the base). You can measure this height using a level line (a string and a clip-on bubble level) and a ruler.

On theCoordinate System Control Panel, select the tracker and set it toAny XY Planeand set the Z coordinate (for Z-up mode), or selectAny XZ Planeand set the Y coordinate (for Y-up mode).

Now you’re ready to go! This setup is a valuable one for outdoor shots where a true vertical reference is required, but the features being tracked are not structured (rocks, bushes, etc).

Again, we recommendnot trying to constrain the camera to be exactly linear, though you can easily set this up by locking Y and Z to be fixed for the duration of the shot, with single-frame locks on X at the beginning and end of the shot. This setup forces the camera motion to be exactly straight, but moving in an unknown fashion in X. Although the motion will be constrained, the setup willnotallow you to use fewer trackers for the solve.

Using a Supplied Camera Path

This section addresses the case where you have been supplied with an existing camera translation path, either from a motion-controlled camera rig, or as a result of hand-editing a previous camera solution, which can be useful in marginal tracks where you have a good idea what the desired camera motion is. After editing the path, you want to find the best orientation data for the given path.

If you have an existing camera path in an external application (either from a rig, or after editing in maya or max, for example), typically you will import it using a standard or custom camera-path import script. Be sure that the solved camera path is cleared first, so that the seed path is loaded.

If you have a solved camera path in SynthEyes, you can edit it directly. First, select the camera, and hit the Blast button on the 3-D panel. This transfers the path data from the solved path store into the seed path store. Clear the solved path and edit the seed path.

Rewind and turn on all 3 camera axis locks: L/R, F/B, and U/D.

Next, configure the solver's seeding method. This requires some care. You can use the Path Seeding methodonly if your existing path includes correct orientation and field of view data. Otherwise, you can use the Automatic method or maybe Seed Points. The Refine mode is not an option since you have already cleared the solution to load the seed path, and don’t have orientation data anyway or you’d use Path Seeding.

You can use Seed Points mode if you are editing the path in SynthEyes—but be sure to hit theSet Allbutton on the Coordinate System Setup Control panelbeforeclearing the prior solution, so that the points are set up properly as seeds. You should probablynotmake them locks, unless you are confident of the positions already.

With the camera path locked to a complex path (other than a straight line), no further coordinate system setup is required, or it will be redundant.

You can solve the scene first with the Constrain checkbox off, then switch to Refine mode, turn on Constrain, and solve again. This will make it apparent during the second solve whether or not you have any problems in your constraint setup, instead of having a solution fail unexpectedly due to conflicting constraints the first time.

Camera-based Coordinate System Setup

The camera axis constraints can be used in small doses to set up the coordinate system, as we’ve seen in the prior sections. Typically you will want to use only 1 or 2 keys on the seed path; 3 or more keys will usually heavily constrain the path and require exact knowledge of the camera move timing.

Roughly, each keyed frame is equivalent in effect to a constrained tracker located at the same spot. You should keep that in mind as you plan your setup, to avoid under- or over-constraining the coordinate system.

Soft Locks

So far we have described hard locks, which force the camera exactly to the specified values. Soft locks pull more gently on the camera path, for example, to add stability to a section of the track with marginal tracking. In either case, for a lock to be active, the corresponding lock button (U/D, L/R, pan, etc) must be on.

The weight values on the Hard and Soft Lock dialog controls whether locks are hard or soft. If the weight is zero (the default), it is a hard lock. A non-zero weight value specifies a soft lock.

Weight values range from 1 to 120, with 60 a nominal neutral value. However, we recommend that when creating soft locks, you start with a weight of 10, and work upwards through 20, 30, etc until the desired effect is obtained.

Weight values are in decibels, a logarithmic scale where 20 decibels is a factor of 10, and 6 decibels is a factor of two. So 40 is 10 times stronger than 20, and 26 is twice as strong as a weight of 20. (Decibels are commonly used for sound level measurements.)

A lock can switch from hard to soft on a frame-by-frame basis, ie frames 0-9 can be hard, and 10-14 soft. You may need to key the weight track carefully to avoid slow transitions from 20 down to 0, for example.

Soft locks are treated as such only when the Constrain check box is on: it is the solver that distinguishes between hard and soft locks. If Constrain is off, the locks will be applied during the final alignment, when they do not affect the path at all, just re-orient it, so soft locks are treated the same as hard locks.

Note that the soft lock weight isnota path blending control. You might naively be tempted to set up a nominal locked path, and try to animate the soft lock weights expecting a smooth blend between the solved path and your animated locked path. But that is not what will happen. The weight changes how seriously SynthEyes takes your request that the camera should be located at the specified position—but it will affect the tracker positions and everything else as well.

Orientation Locks

You can apply Pan, Tilt, and Roll rotation locks as well as translational locks. They can be used for path editing and, to a lesser extent, for coordinate system setup.

For example, a roll-angle constraint can be used to keep the camera forced upright. That can be handy on tripod shots with large pans: small amounts of lens distortion can bend the path into a banana shape; the roll constraint can flatten that back out.

If the camera looks in two different directions with the roll locked, it constrainstwodegrees of freedom: only a single pan angle is undetermined! For example, if looks along the X axis then along the Y axis, both with roll=0. You might want to think about that for a minute.

The perspective window's local-coordinate-system and path-relative handles can help make specific adjustments to the camera path.

Inherently, SynthEyes is not susceptible to “gimbal-lock” problems. However, when you have orientation locks, you are using pan, tilt, and roll axes that do define overall north and south poles, and you may encounter some problems if you are trying to lock the camera almost straight up or down. If this is the case, you may want to change your coordinate system so those views are along the +Y and –Y axes, for example.

Object Tracking

You can also use locks on moving objects, in addition to cameras. However, there are several restrictions on this, because moving objects are solved relative their hosting camera path, but the locks are world coordinate system values.

If a moving object has path locks, then

1.   the host camera must have been previously solved, pre-loaded, or pre-keyed, and the camera solving mode set to Disabled,

2.   the translation axis locks must either all be on, or all off, and

3.   the rotation axis locks must either all be on, or all off.

Normally, when SynthEyes handles shots with a moving camera and moving object, it solves camera and object simultaneously, optimizing them both for the best overall solution. However, when object locks are present, SynthEyes must be able to access the camera solution first, in order to be able to apply the object locks.

With the camera path, SynthEyes changes the translation and rotation axis lock values into a form usable for the object, but the individual axes are no longer available, and either all must be constrained, or none. SynthEyes will automatically turn all the enables on or off together if a moving object is active.

Object locks have very hard-to-think-about impacts on the local coordinate system of the trackers within the object. Most object locks will wind up over-constraining the object coordinate system.

We recommend that object locking be used only to work on the object path, not to try to set up the object coordinate system.