---

Handle mismatched or unconnected points

Following automatic initialization, the next step is to correct the set of initialized 3D points so that they are as accurate and complete as possible. The aim is to create a complete set of 3D points matched in as many camera views as possible as a good starting point for the tracking algorithm.

You need to consider each required point in your marker set and examine it, both in the four 2D camera views, and also in the 3D view. You may need to hand-correct the following classes of error at this stage:

❙       Reconstructed point contains one or more mismatched detections across the four camera views.

❙       A required marker is entirely missing from the set of 3D points.

❙       A required marker is correctly reconstructed in 3D, but additional unconnected detections in one or more camera views would provide useful tracking data.

Reconstructed point contains one or more mismatched detections across the four camera views.

To correct mismatched detections in reconstructed points:

1.    In either a camera view or the 3D view, select the 3D point with the mismatched detection(s).

2.    Do one of the following:

–           Either delete the 3D point entirely by pressing Ctrl+D; or

–           To detach unwanted detections from the point, in the Selection pane, select the channel(s) in which the mismatch(es) occur and then click Edit > Clear Selected [Current Frame].

3.    Before tracking the take, run Bundle Current Frame [Unconstrained].

A required marker is entirely missing from the set of 3D points.

To create a new 3D point:

1.    Selecting matched detections in two or more camera views.

2.    Press M to merge the detections to create a new 3D point.

3.    Before tracking the take, run Bundle Current Frame [Unconstrained].

A required marker is correctly reconstructed in 3D, but additional unconnected detections in one or more camera views would provide useful tracking data.

To add unconnected detections to a point:

1.    In either a camera view or the 3D view, select the existing 3D point (or one of its existing detections).

2.    To add in the additional detections to the point, CTRL+click the missing detection(s) in other camera view(s), and press M.

3.    Before tracking the take, run Bundle Current Frame [Unconstrained].