Accuracy Case Study: Independent Aerotriangulation of a Vexcel UltraCam Dragon Block
Aerotriangulation is the foundation of aerial sensor processing. Every product an aerial campaign delivers — orthophotos, surface and terrain models, point clouds, 3D meshes — inherits its accuracy from this one step: the simultaneous solution of where each camera was, and how it was pointed, at the instant of every exposure. Get it right and accuracy flows all the way to the ground; get it wrong and nothing downstream recovers it. So it is the right place to ask a simple question — how does an independent solution compare to the manufacturer's own?
We took the publicly distributed Vexcel UltraCam Dragon 4.1 demonstration block over Graz, Austria, produced a complete, independent aerotriangulation, and measured its accuracy against the manufacturer's reference solution and against surveyed ground points held back from the adjustment. The headline: the independent solution came in below the commercial reference on image precision, and held survey-grade absolute accuracy against points it never saw.
Accuracy versus the reference
Measured on the same block, the independent solution meets or exceeds the manufacturer's published figures:
| Metric | AG2IIndependent solution | Reference |
|---|---|---|
| Image σ₀ — standard deviation of unit weight | 0.64px | 0.82 px |
| Median image residual | 0.38px | — |
| GNSS position RMS (X / Y / Z) | 0.021 / 0.034 / 0.010m | ~0.050 m |
| Per-camera image residual | 0.59–0.65px | — |
| Boresight magnitudes | < 15mgon | factory |
True accuracy — independent check points
Internal fit statistics flatter themselves. The honest measure is agreement with surveyed marks withheld from the solution — points the adjustment never sees. Against those independent check points, the solution agreed to:
Agreement with the factory camera calibration
As an independent cross-check, the camera's focal lengths — determined from this block alone — were compared against the manufacturer's factory calibration. Every head agreed to within 0.025 mm (< 0.03%) of the factory value, an independent confirmation of the sensor's published interior geometry.
True absolute accuracy, with the report
Anyone can quote a precision figure. Absolute accuracy means agreement with independent ground truth — and the only honest way to claim it is to publish the check-point report behind it. The full case study, with every table above, is available here:
↓ Download the full Accuracy Case Study (PDF, 3 pp)
Study data: Vexcel UltraCam Dragon 4.1 demonstration block — Graz, Austria (publicly distributed). No confidential or client data is involved.