NEW IN R18
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).
There are two ways to solve a 3D object:
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 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 than placed into the Object Tracker’s Geometry Object field (Reconstruction 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:
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 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.
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.
A selected User Track can be renamed in the Name field. The corresponding 3D feature will appear under the same name in the Object Manager after a successful 3D object solve.
Use these settings to define the pattern and search size (only for User Tracks), which work exactly as described under Pattern Size. You can enter separate values if, for example, the object moves much faster than the camera.
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.
Removes selected Tracks for the Object Tracker’s list. They will then be assigned to the Motion Tracker.