< Previous | Contents | Manuals Home | Boris FX | Next >

Object Tracking

Here’s how to do an object-tracking shot, using the example shot lazysue.avi, which shows a revolving kitchen storage tray spinning (called a Lazy Susan in the U.S. for some reason). This shot provides a number of educational opportunities. It can be tracked either automatically or under manual supervision, so both will be described.

The basic point of object tracking is that the shot contains an object whose motion is to be determined so that effects can be added. The camera might also be moving; that motion might also be determined if possible, or the object’s motion can be determined with respect to the moving camera, without concern for the camera’s actual motion.

The object being tracked must exhibit perspective effects during the shot. If the object occupies only a small portion of the image, this will be unlikely. A film or HD source will help provide enough accuracy for perspective shifts to be detected. (If there is little perspective shift, typically a known reference mesh must be used to assist object tracking.)

For object-tracking, all the features being tracked must remain rigidly positioned with respect to one another. For example, if a head is to be tracked, feature points must be selected that are away from the mouth or eyes, which move with respect to one another. If the expression of a face is to be tracked for character animation, see the section on Motion Capture.

Moving-object tracking is substantially simpler than motion capture, and requires only a single shot and no special on-set preparation during shooting.

IMPORTANT : You need to have six or more different trackers visible pretty much all of the time (technically it can get down to 3 for short periods of time but with very low accuracy). Generally you should plan for at least 8-10 to make allowance for short-term problems in some of them. More trackers means more accuracy and less jitter.

Fine Print : no matter how many trackers you have, if they are all on the same plane (the floor, a piece of paper, a tablet, etc), they only count as four, and as the rule says, you must have six! If the object being tracked is flat, you will have to use a known (flat) mesh as a reference, entering known coordinates for each tracker.

Warning : if the object occupies only a portion of the image, it will not supply enough perspective shift to permit the field of view of the lens to be estimated accurately. You must either also do a camera track (even a tripod solve will do), or you must determine a lens field of view by a different method (a different shot, say), and enter it as a Known Field of View.

 

Automatic Tracking Supervised Tracking After Tracking Using a 3-D Model Difficult Situations Goosing the Solver Inverting Perspective Overall Distance Constraints

©2024 Boris FX, Inc. — UNOFFICIAL — Converted from original PDF.