Open topic with navigation

Trackers

Motion Tracker

An Object Tracker needs a corresponding Motion Tracker (otherwise all settings will be grayed out). The link to the Motion Tracker is displayed in this field. If only one Motion Tracker is present in the scene when an Object Tracker is created, this link will be created automatically. Otherwise the currently selected Motion Tracker object will be linked. Normally, this field will be filled out automatically either way (unless you experiment with several Motion Tracker objects at once and want to determine yourself which one is linked).

Geometry Object

There are two ways to solve a 3D object:

Geometry-assisted Solve

With this method, a 3D object is modeled (referred to as ,reference object’ in the following) that will follow an object in the live footage as precisely as possible. This object should be well marked with 2D tracks to make it easier to track. This can - but doesn't have to be - the same object that will be replaced by the 3D object in the footage (it should, however, be closely related). Assuming you want to replace a person's thumb with a robot thumb. You can track the thumb, which would be difficult because it has very few visual characteristics that are easy to track. Instead, you can use a checkered pattern (in the real world you would use a special tracking glove in the live footage) attached to the thumb that moves as the thumb does.

A congruent plane was positioned over the checkerboard pattern.

A plane was created (make a note of the frame; later the reference object must be made a Child object of the Object Tracker at this frame if you want it to move with it) as a reference object and positioned to lie over the checkerboard pattern. The reference object has the same size as in the live footage (in this example 14 cm x 14 cm, for example).

This layer is then placed into the Object Tracker's Geometry Object field (Tracker tab). The following must be true for the geometry object:

Note that a focal length must be defined for the Motion Tracker object (Reconstruction tab) when using this method.

The following happens when you click on the Run 3D Solver for Object button:

  1. Rays will be emitted, at the time in the footage as specified above, into the scene from the camera's angle of view and the points of intersection will be calculated with the reference object.
  2. These points of intersection can be used to assign 3D positions to 2D Tracks in the object space.
  3. Position and rotation of the reference object in relation to the camera can be calculated for other segments in the footage using a combination of 2D Tracks and known 3D positions.
  4. Keyframes will be created for the Object Tracker to give the object the correct movement in relation to the camera (which themselves may already be animated via a previous camera solve). If the solve was successful, the matching 3D object must be made a Child object of the Object Tracker so it moves correctly over the entire length of the live footage. This is the best way to see if the 3D object solve was successful and if the quality is high enough.

The geometry-assisted solve requires fewer Tracks than without geometry. Many times, objects tracked in the live footage are very small and only few Tracks are needed. The absolute minimum number of Tracks for this method is 3 for each point in time (if you have 3 good Tracks (that must also lie on the geometry object from the angle of view of the camera), a simple plane can suffice as a reference object to create a 3D solve). More Tracks - in adequate quality - will, however, lead to more precise results.

By using a polygon object as a reference object, the scene will also be calibrated, i.e., you will not require a Motion Tracker Constraint tag (which is not required in conjunction with the aforementioned method), contrary to the geometry-free method described below.

Geometry-free Solve

Geometry-free solves (the Geometry Object field remains empty) creates a group of points for the Tracks (absolute minimum is 7 Object Tracks; the more Tracks with adequate quality, the better) that will initially lie in space with no defined scale or distance from the camera; this is similar to Camera Tracking as long as its solve has not yet been ,calibrated’ by a Motion Tracker Constraint tag.

This only means that the undefined group of points (which is first of all independent of any camera solve) must first be ,calibrated’ using the Motion Tracker Constraint tag. The axes and positions that you define here refer to the Object Tracker's coordinate system (contrary to camera solving which only uses the world coordinate system)! This means that when you create a polygon object to match the movement and make it a Child object of the Object Tracker, its position and rotation will be set to 0 so it is set to the Object Tracker's origin from where it will move along correctly.

Assigned Tracks

Tip:Note that the Track settings defined here also apply to the Motion Tracker Object's settings (and vice-versa).

All Tracks (Auto and User Tracks) from the tracked object are displayed in this list. This lets the Motion Tracker object recognize which Tracks should be used for camera tracking: all, except for those shown in the list.

The selections made here correspond with the selections in the Viewport.

The description of the list can be foune here in the Object Tracker object. The functionality is exactly the same.

Virtual Keyframes

This option enables or disables virtual keyframes for selected Tracks.

Normally, a User Track is created by creating a key and tracking it. This tracking uses the key saved to the pixel pattern, which, of course, changes over the course of the video as it moves farther from the frame. If the change is too great, the tracking will fail and you have to make adjustments manually. Virtual keyframes help with tracking by automatically adding new keys to the tracked section according to the settings below (contrary to manually-created keyframes - which are permanent - these virtual keyframes are dependent on the virtual keyframe settings, among other things, and can therefore change).

This will produce better tracking in most cases since changing pixel patterns can be compensated for by the virtual keyframes. However, it can happen that the virtual keyframes can shift from their original, manually defined footage position. If this happens, keys can be added manually.

Virtual keyframes are colored light blue in the Motion Tracker graph.

Mode

Here you can define the rules according to which virtual keyframes should be created. The following options are available:

Error Threshold[-∞..+∞]

Internally, the difference to the previous key (corresponding to the Tracking direction) will be ascertained for each Track at each frame. If this difference is greater than the threshold value defined here, a virtual keyframe will be created. The smaller the value, the less modification that will be allowed compared to the previous frame and the more virtual keyframes will be created and vice-versa.

Frame Interval[-2147483648..2147483647]

Use this setting to define the temporal interval in which virtual keyframes should be created.

Color Filter

Filter.Red[-200..200%]

Filter.Green[-200..200%]

Filter.Blue[-200..200%]

This setting can be used to change the color weighting of selected Tracks. A new Tracking will be created immediately. Details about color weighting can be found under Preferences: Weighting Red/Green/Blue.

In the Color Filter selection menu you can select from the following options:

Assign Selected

If Tracks are selected in the Viewport, they can be added to the list using this command. This can also be done per RMB: Assign to Object.

Unassign Selected

Removes selected Tracks for the Object Tracker's list. They will then be assigned to the Motion Tracker.