Getting Better Results with the Planar Tracker

You can use planar tracking in your Action compositing workflow on many different types of footage to produce different results. The tracking menu exposes many settings to help you prepare and refine the planar track. Read the following sections for some tips and best practices to get the best possible results out of the planar tracker.

What Result Are You Looking For?

Before you perform planar tracking, it is useful to think of what type of result you are trying to achieve in Action. Since you can apply planar tracking from the Axis, Perspective Grid, or GMask node, you should have an idea of how each node handles the resulting tracking data:

Planar Tracking in the Axis node: Use when you want another object or surface to follow to transform happening in an images sequence. Since the Axis itself does not contain media, the planar tracker needs GMask splines to include or exclude to perform the tracking analysis.
Planar Tracking in the Perspective Grid node: Use when you have images with planar info that you want the perspective to be sorted out for you. Since the Perspective Grid acts as an Axis, it performs a similar tracking function, but has the advantage of tracking the grid information directly (with or without GMask constraints).
Planar Tracking in the GMask node: Use primarily for rotoscoping work (to remove parts of the image or the Action scene for another tool or compositing), or if you want to use a GMask as an animated occlude.

Pre-Tracking Settings

See the full list of settings here or in the tooltips. Here are a few things to keep in mind before you track:

Look out for rotation, scale, skew, and perspective in your media to be tracked; and enable the appropriate buttons to help the tracking analysis. Of these four buttons, pay particular attention to Perspective; enable it only if the content is planar, and there is perspective information. Enabling Perspective without these may result in instability in the tracking analysis.
Sampling field (Region Warping algorithm): As a general rule, a larger sampling reference increases the quality of the results, but takes longer. There may be instances, though, when a lower sampling value works better, such as when there is a lot of grain or noise in the media, or if there is fast camera motion. In these instances, down-sampling may actually eliminate these potential sources of confusion for the algorithm.
Lighting button (Region Warping algorithm): Enable to help with colour variations or high specularity. Enabling Lighting can also help with diffuse reflection in an object, such as clouds in a window.
Occlusions button (Region Warping algorithm): Make sure that the sequence has occlusions that you don't want to be tracked. Enabling on a sequence where no occlusions appear, may lead to tracking errors.
Analyze RGB Channels button (Region Warping algorithm): Should only be enabled when colour information can make a difference in a track. When enabled, it could generate a better result than when only Luminance is used, but at the cost of performance.
In some cases, bad tracking data can be alleviated by selecting Lock in the FOV Adjustments box in the Perspective Grid menu, before tracking.

Analysis Constraints

Use inclusive or exclusive GMask constraints to limit the tracking to certain areas of the media. When planar tracking from the Axis node, you must set GMask constraints, as the Axis itself can not be tracked.

Inclusive constraints: You can set GMask constraints (which include GMask gradients, if created), or just GMask gradients (inner and outer gradients can be tracked). An example of using only GMask gradients is when you are tracking a highly reflective surface, such as a smart phone screen. In this case, you can create a gradient on the GMask to expand the trackable area.
Note: When tracking from the GMask node, Gradients is the only available inclusive constraint.
Exclusive constraints: You can set areas of the media to not be considered in the tracking. Media GMasks float in the Action scene (that is, they are not attached to the node being tracked), but they must be assigned to the same media as the node being tracked.

Analysis Settings

To help you view the tracking analysis as it occurs, switch to Object Solo (F8) view, and make sure that the Shape Reference and Shape Current buttons are enabled in the Display tab.

If you need to fine-tune the tracking analysis, many of the analysis settings are similar to those in the Stabilizer; Step, Snap, and Auto Update Reference should be familiar to you, as well as using the backspace key to delete bad keyframes (it is preferable to delete any bad keyframes before re-tracking). See the full list of settings here or in the tooltips.

Use Ctrl for added functionality with the Go To Reference button to go to the last analyzed frame, and with the Set button to add a keyframe on the track without doing an analysis.
Updating the reference can help the track, since it constantly compares the current frame content to the reference, to extract the tracking data. You have two options to update the reference:
- Update it manually when the track is failing, using Snap.
- Enable Auto Update Reference. Using Auto Update Reference generally allows you to perform an easier track, at the cost of stability, since it compares the current frame to the previous (or next) frame, and therefore is likely to introduce offset over time, or instability in the track. Large occlusions or motion blur may also impede the Auto Update Reference.