|PFTrack Documentation||Node Reference|
The Auto Match node can be used to automatically generate trackers to use for camera or object solving, and has a single input and output.
The Auto Match node attempts to pick points in each frame and match them together to form a set of connected feature tracks. In many situations, this matching process can be performed more reliably than when attempting to track individual features between frames using the Auto Track node, and results in a more accurate set of feature points with fewer errors and less drift over longer clips.
The Auto Match node is designed to work best with features picked from natural images, such as marks on a wall, or stone or fabric texture for example. For scenes containing more synthetic features such as tracking markers on a green-screen set, the Auto Track or User Track nodes may provide better results.
All trackers generated by a single Auto Match node are placed into a tracking group. By default, this group is called Camera and corresponds to tracking the overall camera motion. When tracking an object, the group should be changed to something representative of the object being tracked (for example, "Car" or "Head").
The Auto Match node works by first identifying a large number of candidate points in each frame, and matching those points from frame-to-frame over the entire clip. Once the candidate points have been matched, they are filtered down into a smaller set of tracking points which are then passed down-stream.
The tracking points are chosen so that they best represent the overall contents in the scene, exist for a large number of frames, and are well distributed over the image area. The number of tracking points can be adjusted interactively without the need to re-generate all the candidate points each time, greatly speeding up the process of selecting suitable parameters for a shot.
Candidate points in each frame are selected by looking at the image data and identifying areas that correspond to strong features such as corners or image texture. The number of candidate points that are then selected for matching in each frame is controlled by two main parameters:
- The Feature Scale value determines the overall scale of features that examined. When this is set to Small, small corners and image texture regions will be used. Large means larger scale features will be selected. Looking for features at a larger scale can be beneficial in situations where image noise or other factors have a detrimental effect on the selection of candidate points.
- The Feature Strength value specifies the fraction of points that are passed on to the matching process. For example, a Feature Strength of 75% means the strongest 75% of features are kept for matching, and the weakest 25% are rejected.
Masks can be generated to restrict candidate point selection to a specific part of the image. This is essential when the clip contains objects moving independently from the camera, and points are being generated for either camera or object tracking.
When matching candidate points from one frame to the next, the Matching Prediction and Matching Area parameters are used to define a search area over which points are compared. If the clip exhibits short bursts of motion that are inconsistent with the Matching Area values, all point matches can be lost. For these cases, a custom Matching Area can be set by adjusting the values in the Matching Area edit boxes.
The statistics graph displays information about the number of tracking points that have been generated for the clip. The graph displays a colour-coded count of the number of tracking points in each frame.
The graph is colour-coded according to the length of each tracking point. Points that exist for 5 frames or less are coloured red, and points that exist for more than 50 frames are coloured green, with graduations of orange and yellow between.
Below the graph is a numerical breakdown of the overall length of each tracking point, including the average track length, and a percentage breakdown of the entire set of tracking points, colour-coded using the same scheme as the graph.
The Coverage Panel displays information about the frames in which each tracking point has been tracked:
This can be used to evaluate how well tracking points are distributed throughout the clip, which will help to provide an accurate camera solve without any jumps in the camera path.
The Coverage Panel displays each tracker as a horizontal chart, illustrating the frames in which each tracker is visible. The current frame is displayed using white vertical lines.
Each frame in which the tracker is present is indicated by a blue square. Lighter-blue squares indicate frames in which the tracker was first located. Clicking on an indicator will select the tracking point and display that frame in the Cinema window. Holding the Ctrl key will allow multiple tracking points to be selected.
The panel can be panned horizontally or vertically by clicking and dragging with the right mouse button. Clicking and dragging with the middle mouse button will zoom either horizontally or vertically to increase the number of tracking points and frames that are displayed in the panel.
The mouse wheel can also be used to zoom horizontally, or vertically if the Alt/Option key is held.
Additional controls are also available:
All: Switch between displaying all trackers, or only the trackers visible in the current frame.
Name: Sort the tracking points by name, in alphabetical order.
Start: Sort the tracking points according to the first frame in which they are tracked.
End: Sort the tracking points according to the last frame in which they are tracked.
Fit: Fit the tracking points display to the window. This will zoom in or out as necessary, displaying as many tracking points and frames as will fit in the viewport.
Create a new tracking group.
Current group: Specify the group that new tracking points will be associated with. Groups can be renamed by typing into the menu box.
Delete the current tracking group.
From: Set the first frame to use for point matching to the current frame. The corresponding frame number is displayed in the edit box. Reset with R button.
To: The last frame to use for point matching to the current frame number. The corresponding frame number is displayed in the edit box. Reset with R button.
Target number: Target number of tracking points to generate for each frame.
Min length: The minimum number of frames that each tracking point must exist for.
Reset: Reset the point parameters to their default values.
Store current parameters as a new preset.
Current preset: Current preset for the matching parameters. The current preset name can be changed by typing into the menu box.
Delete the current preset.
Channels: Image and mask channels to use for point selection and matching.
Image proxy: Image resolution to use for matching. The default value (None) will track the image at full resolution. Other options are Half, Third and Quarter, to track the image at half, third or quarter resolution.
Feature scale: Scale of features expected in the image. When set to Large, only large-scale corners and image features will be picked. Using a Small feature scale (the default value) can sometime mean noise or image grain is mistaken for a significant corner, especially when using high resolution images.
Feature Strength %: The number of candidate points that are matched in each frame, based on their relative feature strengths. The default value is 50%, meaning that the 50% of strongest features will be matched, and the rest rejected.
Matching area: The width and height of the search area (in pixels) used when attempting to match candidate points in one frame to another.
Matching prediction: The method used to predict where one candidate feature should be matched to in the next frame. The default value is Image Motion, indicating that the average motion seen over the 2D image should be used. Alternatively, setting this to From Metadata will use metadata describing approximate camera motion if that is available for the clip. Setting this to None means that no prediction is used.
Reset: Reset the tracking parameters back to their default values.
Skip Frame: Skip the current frame during feature matching. Once skipped, the button label will change to Un-Skip Frame to include the frame during the matching process. Skipping frames can be useful when one frame is missing or corrupted due to some sort of image degradation.
Edit ROI: Adjust the region of interest (ROI) for matching in the Cinema window. Click and drag with the left mouse button to adjust edges of the ROI. Points outside the ROI will be rejected during the matching process.
Auto-Match: Start the automatic matching process, matching candidate features throughout the clip and automatically filtering them to generate a set of tracking points. Holding the Shift key whilst clicking this button will run the matching process in the background.
Preview Match: When enabled, candidate point locations will be displayed in the Cinema window. This can be used to quickly assess how different parameters affect the features before matching. Hovering the mouse near a preview point will display a numerical readout of the tracking motion to the next frame.
Preview search windows: When enabled, hovering the mouse over a point will also display the Search Window that was used to match the point into the next frame.
Preview failures: When enabled, candidate points that were not matched are also displayed. In this case, if a point was not matched because it was not considered to be a strong enough feature, it is displayed in light blue. These are points that have not passed the Feature Strength test, described above.
Show Paths: When enabled, all point paths will be displayed in the Cinema window.
Show Info: When enabled, pixel coordinate information will be displayed in the Cinema window.
Mouse controls can be customised in the Preferences.
Individual tracking points can be selected in the Cinema window using the left mouse button (holding the Ctrl key for multiple selection). Once selected, press the Del key to delete the tracking points.
Display a tooltip showing the statistics for the current frame displayed in the Cinema window.
Zoom the statistics graph.
Pan the statistics graph.
Keyboard shortcuts can be customised in the Preferences.
|Preview Search Windows||Ctrl+T|