How does motion tracking work?

The tracking behaviors in Motion analyze an area of pixels known as a reference pattern over a range of frames in a movie clip to “lock onto” the pattern as it moves across the Canvas. You specify the reference pattern to be analyzed and “tracked” by dragging one or more onscreen trackers (a red crosshair in a circle, shown below) to the area of the clip you want to analyze. Motion then analyzes the motion of the designated reference pattern for a specified duration of time. This duration is based on the length of the tracking behavior, the length of the defined play range, or the length of the clip.

Ideally, the reference pattern should be a consistent, easily identifiable detail with high contrast. This makes the pattern easier to track.

During the analysis, the tracker advances to each subsequent frame, sampling many positions in the search region around the center point of the tracker. Some of those positions fit the designated reference pattern more closely than others; the tracker finds the position where the search region most closely matches the reference pattern (with subpixel accuracy). For every frame analyzed, the tracker assigns a correlation value by measuring how close the best match is.

In addition to searching for the reference pattern’s position, the tracker identifies how the pattern transforms (scales, rotates, or shears) from one frame to the next. Imagine you’re tracking a logo on the shirt sleeve of a person walking past the camera. If the person turns slightly as he passes the camera, the reference pattern rotates. The tracker looks for the reference pattern and any shifts in that pattern’s scale or rotation.

As it analyzes motion in your project, Motion records the data, which you can then apply to any other object in your project.

The more trackers you use, the more spatial information you’ll record: One-point tracking records position data; two-point tracking and four-point tracking record position, rotation, and scale data (by comparing the relative change between the points); multiple-point tracking can record all the control points (vertices) on a shape.

Tracking in Motion is planar; in other words, tracking does not occur in Z space (depth). For example, when you analyze two features in a clip—and that clip is moving in 3D space—you record the changes in position, scale, or rotation over time in the clip, but not its actual 3D transformation.