The Tracker first tracks motion of the reference
feature(s), then applies or inverts that motion, depending on if
you are tracking or stabilizing.
Once you have set up
your trackers, the analysis can begin. This is how the Tracker analyzes
and remembers the motion of the reference feature:
- The Tracker stores the position of the
reference box on frame 1.
- The Tracker advances to the next frame
and positions the tracker box in the same place as in the previous
frame.
- It searches the tracker box for an area
that matches the contents of the reference box. The Tracker uses
luma cues (white versus black) and shape cues to determine what
is a match.
- Once it finds a match, the Tracker uses
the location of the match to move the center of the tracker box
over the center of the match. This change in position is stored
as an X and Y shift value.
- On the next frame, and on all future
frames, the Tracker gets a little smarter about where it positions
the tracker box. It extrapolates the likely position of the reference
feature based on the amount and direction of the motion that occurred
between the last two frames. The Tracker makes an educated guess
of where the feature may be, based on past matches.
NoteThis means the Tracker
works best when the motion of the reference feature is regular and
does not suddenly or erratically change direction or speed.
- Once the Tracker has analyzed the motion
of the reference feature, and stored it as Shift values for each
frame, it copies the motion of the reference feature to the element
that you are tracking, so that the two appear locked together.
In all tracking except
for stabilizing, the Tracker applies the shift values without modification.
With stabilizing, the
Stabilizer operator inverts the shift values, resulting in a negation
(and therefore elimination) of the motion—see
Stabilizing.