Skip to content
ART — Advanced Realtime Tracking
GermanEnglish
← Blog & Guides & Papers

Tracking quality dropping? → Room calibration first — body calibration only when needed

When DTRACK no longer tracks reliably: why room calibration or room recalibration should come before a new body calibration — with 2D Monitor examples.

  • DTRACK
  • Calibration
  • Room calibration
  • Body calibration

“It’s not working properly any more!”

Users of camera-based tracking systems know the situation: the optical system was installed and calibrated, worked well for a long time — and in recent weeks it no longer feels as good as it used to. Targets take longer to acquire, tracking dropouts occur more often, and working with the system is no longer satisfying.

The first step that users try is often a new body calibration. Here we explain why a different approach is usually more productive.

(This guide applies to DTRACK3 and DTRACK4 with ARTTRACK5, ARTTRACK6/M, and AT7-50/80 single cameras, as well as SMARTTRACK.)

How can I spot tracking problems — beyond gut feeling?

Tracking dropouts show up in the DTRACK GUI colour code (red/green) and usually in the application itself. In DTRACK it is helpful to take a closer look on the 2D Monitor of the individual tracking cameras: besides marker positions, each camera window shows how many markers are used for the 6DOF calculation:

DTRACK 2D Monitor: three cameras, each 5/5 — room and body calibration are good
DTRACK 2D Monitor: three cameras, each 5/5 — room and body calibration are good

The 5/5 values in the headlines mean: each of the three cameras sees five markers (first number), and DTRACK uses all five markers from all cameras for 6DOF calculation (second number). Room and body calibration are in good shape.

These 2D Monitor numbers often let you distinguish bad room calibration from deformed targets — see the examples below. When tracking troubles you, the 2D Monitor should be your first stop.

Additionally, DTRACK can optionally output 6DOF covariances to detect quality issues before tracking dropouts occur. Details are in the white paper Using 6dof covariances for online assessment of tracking quality — available from ART on request and coming soon in this blog.

What are the most frequent reasons for bad tracking?

Besides external or setup-related causes — sunlight reflections, occlusions, worn markers — there are three calibration-related factors:

  1. External orientation (EOR): The position and angle of one or more cameras has changed since the last room calibration or room recalibration.
  2. Internal orientation (IOR): Camera calibration parameters have changed since factory calibration.
  3. Body geometry: The target geometry no longer matches the data from the last body calibration.

In practice, cause 1 is by far the most common. External orientations — camera positions and mutual alignment — can shift for many reasons: aluminium trusses expand with temperature, whole buildings move between summer and winter, ceiling mounts drift. Stress in supporting structures can cause sudden movement — as can cleaning staff or tradespeople working after hours.

SMARTTRACK is factory-calibrated and self-contained: if the whole unit moves, calibration stays valid. Still, camera heads can shift slightly with temperature or heavy acceleration during transport — especially on older SMARTTRACK2 units. SMARTTRACK3 is better suited to rough operating conditions.

Two orders of magnitude for context:

  • In a multi-camera system with a 5 m camera-to-target distance, a camera rotation as small as 0.05° can stop that camera from contributing to tracking.
  • On SMARTTRACK, a relative rotation of the camera heads of 0.15° or more can reduce the maximum tracking distance to 1.5 m or less.

Critical changes to IOR or target deformation occur much less often.

On camera calibration: We often see 10-year-old ART cameras return from the field and still show sufficiently good IOR in our lab — IOR recalibration is often unnecessary, though it does no harm.

On rigid bodies: Most targets in use — especially 3D-printed ones — break rather than deform. Deformations usually come from deliberate changes by the customer; as long as the rigid body stays the same, no new body calibration is needed because geometry data is stored in DTRACK.

As stated above, the cameras’ 2D Monitor can give you more hints. Here are a few examples:

2D Monitor: camera c06 delivers 5/0 — room calibration for c06 is invalid
2D Monitor: camera c06 delivers 5/0 — room calibration for c06 is invalid

Cameras c05 and c07 each deliver 5/5; c06 sees five markers but uses none for 6DOF. This means: c05 and c07 are still well aligned; room calibration for c06 is no longer valid — room calibration or room recalibration is required. Because c05 and c07 use all markers, there is no target deformation.

2D Monitor: c05 only 5/4 — room calibration more likely than a body problem
2D Monitor: c05 only 5/4 — room calibration more likely than a body problem

Less obvious: for c05, only four of five detected markers are used; the other cameras show 5/5. c05 may be somewhere between “well aligned” and “completely misaligned” — or one marker may be partially occluded. Here too, deteriorated room calibration is more likely than deformation.

2D Monitor: all cameras 5 markers, but only 4 markers used for 6DOF — check body calibration
2D Monitor: all cameras 5 markers, but only 4 markers used for 6DOF — check body calibration

A rare body calibration issue: all three cameras see five markers, but only four are used for 6DOF. One marker position does not match the geometry stored in DTRACK — due to deformation or a bad previous calibration where only four markers were captured. A new body calibration makes sense here.

Step-by-step procedure

When tracking is poor, we suggest:

  1. Check 2D Monitor — keeping the examples above in mind.
  2. Always perform room calibration or room recalibration next — even if you suspect body calibration (you need a fresh room calibration to recalibrate the target anyway). Room recalibration without the calibration angle does not change the coordinate system and is preferable when cameras have not moved much.
  3. Bring the target into the tracking volume and test. If it works well now, you are done. If not — or if step 1 pointed to body calibration — continue with step 4.
  4. New body calibration using the DTRACK target library or custom body calibration. For custom body calibration:
    • In the result window, always check that all markers of the rigid body were detected.
    • Check that detected marker positions are realistic (phantom markers may falsely appear where no real marker is located).

Remarks:

  • Custom calibration of multi-sided targets (e.g. HMD targets) is often not possible on “small” systems (SMARTTRACK only, or two to four cameras). Contact ART — a lab calibration at the ART office may be necessary.
  • For SMARTTRACK, ART offers the Calibration Checker: check room calibration quality with a wand movement and distinguish EOR from IOR quality. Not in the download section — request it from your ART representative.

Next steps