Patch Match Threshold

This value sets the threshold at which a potential match (either between two detections, or two patches), is considered to be a true match. At a value of 1, each frame must perfectly match; at a value of 0, almost anything is a match.

The default is 0.8.

Face Point Tracking Distance

This value determines how far (in pixels) CaraPost looks from one frame to the next when performing blob matching (or patch tracking) for marker points on the face.

The default is 35 pixels. The higher this value, the more facial movement CaraPost can handle, but increasing this value may also result in more mismatched points. For sequences involving normal facial motion (that is, with no extreme expressions or big changes), if you need to reduce the number of mismatched points, turn on Outlier Removal Threshold 3D (see below). If you still have too many mismatches, decrease the Tracking Distance value.

Face Point Max Elastic Prediction Distance (pixels)

This value determines how far (in pixels) CaraPost looks from one frame to the next when looking for matching blob (or patch tracking), for marker points on the face, after some initial blob matches have been made and Elastic Prediction techniques can be used to estimate a more accurate position for each tracked facial marker point.

The default is 15 pixels, and in general, you should set this value to be smaller than that of the Face Point Tracking Distance. The higher this value, the more facial movement CaraPost can handle, but increasing this value may also result in more mismatched points.

Brim Point Tracking Distance

This value determines how far (in pixels) CaraPost looks from one frame to the next when looking for an initial matching blob (or patch tracking), for marker points on the brim.

The default is only 15 pixels, as the brim is intended to be relatively fixed from one frame to the next. The higher this value, the more movement of the brim CaraPost can handle, but increasing this value may also result in more mismatched brim points.

Patch Match Size

This parameter controls the size of the area that is used to compare potential matches between two frames, when matching detections.

The default is 15 pixels, which means that patch-tracking searches for patches of 15x15 grayscale pixels that are centered on a detection. If you increase this value, fewer outlier matches are found, but the overall number of matches may be reduced.

Remove Bad 2D Projections

This toggles whether matched detections are removed from reconstructed 3D points based on their 2D reprojection error. Enabling this setting reduces the number of mismatched points.

The default is selected (that is, set to True).

Max Allowed 2D Projection Error

A 2D reprojection error is the difference between the position of a 3D point when it is projected back onto the 2D image, and the center of the 2D detection in the image. The Max Allowed 2D Projection Error defines the maximum allowed value for 2D reprojection error. If Remove Bad 2D Projections is enabled, all detections with reprojection errors above the threshold are detached from the corresponding 3D point.

If you increase the value of Max Allowed 2D Projection Error, fewer detections are detached (and corresponding 3D points removed) due to being regarded as mismatches, but a greater number of mismatches may occur.

For takes containing a lot of movement or dense clouds of markers, ensure Remove Bad 2D Projections is selected (set to True) and adjust Max Allowed 2D Projection Error to obtain the most accurate results.

The default is 5 pixels.

Remove Bad 3D Points

If selected, this option further reduces the number of mismatched points/detection pairings by detaching all detections that cause a 3D point to be reconstructed in a location which is inconsistent with the location of neighboring points in the marker set.

Its use results in cleaner data, but it slightly increases the processing time.

The default is selected (that is, set to True).

Max Allowed 3D Point Error

This value specifies the distance that a 3D point has to be from its expected position in 3D, estimated based on the locations of its neighboring points. 3D points that are found at a distance greater than this value from their expected position are regarded as ‘bad’ points and detections that cause the incorrect position of a point are detached from the point.

If you increase the value of Max Allowed 3D Point Error, fewer detections (and corresponding 3D points) are removed due to being regarded as bad points, but a greater number of mismatches may also occur.

As for 2D removal of bad points, for takes containing a lot of movement or dense clouds of markers, ensure Remove Bad 3D Points is selected (set to True) and adjust the Max Allowed 3D Point Error value to obtain the most accurate results.

The default is 5mm.

Allow Single Sensor Tracking

If selected (set to True), CaraPost enables the single sensor tracking option during the tracking algorithm. This feature allows an estimate of 3D point positions to be made based upon a correspondence between a 3D point and a detection in a single sensor (or camera) in certain cases, by constraining the 3D point to lie on the ray associated with the single detection. Enabling this feature can allow CaraPost to track points in certain cases where they are occluded in three of the four camera views, resulting in better continuity of tracking across the take. As long as the Single Sensor Tracking Threshold (see below) is set sufficiently tightly, enabling this feature should not decrease the overall accuracy of the tracked 3D points.

With this option cleared (set to False), single-sensor tracking is disabled.

Single Sensor Tracking Threshold

This value specifies the maximum allowed 3D distance (in mm) that a 3D point calculated using Single Sensor Tracking can be from its expected position in 3D, estimated based upon the locations of its neighboring points. 3D points calculated at a distance greater than this value from their expected position are not included as Single Sensor Tracking results.

The default value for this threshold is 0.5mm, and the value should be set no higher than the maximum tolerated error in mm of the final 3D point positions.

Blob Size Difference Threshold

This value specifies a threshold which determines how similar a blob must be in width and height from one frame to the next to be considered a possible match. The value varies from 0 (no difference in blob width or height allowed) to 1 (any difference in blob width or height allowed).

The default value for this threshold is 0.75, which allows quite a large frame-to-frame difference in blob width or height to be considered a possible match. In cases where you get markers occluding each other, resulting in merged blobs, (for example a hemispherical marker occluding a flat marker), it may be useful to reduce the threshold to around 0.1 so that the merged blobs are discounted as possible blob match candidates.