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

6DOF covariances in DTRACK: monitor tracking quality online

DTRACK can output 6DOF covariances — how to enable the option and use reference values to detect calibration-related quality loss before tracking fails.

  • DTRACK
  • Calibration
  • Tracking quality
  • Motion Tracking

Detect tracking quality early — not only when tracking fails

When tracking quality drops, calibration-related causes are often involved: shifted external orientation (EOR), changed camera parameters (IOR), or outdated body calibration data. EOR changes in particular — from temperature, building movement, or slightly shifted mounts — can develop gradually over days or weeks.

For some applications, waiting until tracking interruptions or recognition problems appear is not acceptable. DTRACK can optionally output 6DOF covariances in the data stream — a way to spot deterioration before visible tracking issues occur.

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

Background on typical calibration causes and the 2D monitor: Tracking quality dropping? → Room calibration first.

Enabling covariance output

DTRACK3

Output of 6DOF covariances requires a complimentary license available from your ART representative. After activation, the option appears in menu Tracking → Output. For 3D markers, 3DOF covariance output can be enabled as well.

DTRACK4

In DTRACK4, 6DOF covariance output is available without an extra license — also under Tracking → Output, together with 3DOF covariance for 3D markers.

DTRACK menu Tracking → Output: 6DOF and 3DOF covariance output options

Basic approach: build a reference database

Covariance values describe the statistical uncertainty of measured 6DOF data. They become useful when compared against reference values:

  1. Right after a careful new calibration (room calibration, and body calibration if needed), capture covariance data.
  2. Build a database of reference covariances for one or more regions of the measurement volume — using the same rigid bodies that will be tracked in operation.
  3. During operation, compare covariances from the DTRACK data stream with the reference values.
  4. If current values in the same volume (and for the same target) are considerably worse than the reference, this indicates a calibration-related issue — you can warn users before the application fails.

How the reference database is structured depends strongly on the specific application. DTRACK cannot provide a general assessment logic — you implement the evaluation in your application.

White paper with details and sample code

A detailed description, interpretation of covariance matrices, and sample code for programmers are in the white paper:

DTRACK3 & DTRACK4 — 6DOF Covariances (PDF)

Further reading