Skip to content

Camera stations: when to use them and the nodal-head requirement

  • Status: unverified
  • Applies to: Metashape Pro 2.x ; Metashape Standard 2.x — and unchanged from PhotoScan 1.x
  • Edition: Standard
  • Diátaxis: explanation
  • Confidence: high
  • Last reviewed: 2026-05-28

Confidence: high. The nodal-head requirement is directly attested by Agisoft support with permalink. The Camera Station group type and its usage are documented in the official manual (Camera station data section).

A camera station in Metashape is a group of images that the software treats as having been captured from the same physical optical centre, with only orientation varying. Photogrammetric processing handles them differently from independent cameras.

What a camera station is

The official manual describes the feature: a CameraGroup with type Station instructs Metashape that all images in the group share one optical centre. Bundle adjustment then solves for one position plus N orientations, instead of N independent 6-degree-of-freedom poses.

Demo verified: ✗ — pending Tier 3 reproduction on a real Metashape install.

import Metashape

chunk = Metashape.app.document.chunk

group = chunk.addCameraGroup()
group.label = "Station 1 (tripod, north view)"
group.type = Metashape.CameraGroup.Type.Station
for cam in chunk.cameras[:5]:   # the 5 images shot from this station
    cam.group = group

After this, matchPhotos and alignCameras respect the station constraint: tie points within the group are skipped (no parallax), and the group's images are placed in 3D space with a single shared centre.

The hidden requirement: nodal-head capture

What the manual does not emphasise: this only works correctly if the images really were captured from the same optical centre. Metashape does not detect or correct sub-millimetre translation drift between station images. Hand-held capture introduces such drift; a properly-set tripod head with the camera rotated about its lens nodal point does not.

"PhotoScan assumes that all images in the camera station group are taken from exactly the same camera position, so using nodal head is recommended. Offsets would not be compensated and may lead to the reconstruction problems." — Alexey Pasumansky, 2015-03-27, PhotoScan 1.1 (permalink)

Practical consequences:

  1. A regular tripod head is not a nodal head. The camera rotates around the tripod ball-mount, not around the lens nodal point. This introduces translation as the camera swings.
  2. Hand-held "panorama-style" images are NOT camera stations. Even careful spinning in place introduces decimetre-scale translation drift between images.
  3. Bracketed exposures from a fixed tripod ARE camera stations (no orientation change either, which means the bundle has nothing to solve within the group; the group simply prevents redundant tie-point matching across the bracketed images).

When to use camera stations

Capture pattern Use station?
Spherical / 360° panorama from a nodal-head tripod ✓ One station per pano
Bracketed exposures (HDR) from a tripod ✓ One station per location
Multi-shot focus-stacking from a tripod ✓ One station per stack
Hand-held "photogrammetry-style" coverage ✗ Each image is its own
Drone flight (each image at a unique position) ✗ Each image is its own
Tripod-mounted camera with deliberate translation ✗ Each image is its own

Implications for 3D reconstruction

A single camera station, by itself, cannot reconstruct 3D geometry. With one optical centre, there is no parallax — only orientation. Reconstruction requires at least two stations with a non-zero baseline between them.

The official manual states this explicitly: photogrammetric processing requires at least two camera stations with overlapping images (Metashape Pro User Manual 2.3, § Camera station data, p. 39).

This is a feature, not a bug: the station type tells the bundle "don't try to extract depth from within this group; treat it as a single observer". The depth comes from the inter-station baseline.

The 360° panorama use case

The most common legitimate use of camera stations:

  1. Mount the camera on a nodal-head tripod, rotate to the nodal point.
  2. Capture multiple images per station, covering the full sphere or the relevant azimuth range.
  3. Translate to a new station (typically 1–10 m baseline for close-range scenes), repeat.
  4. In Metashape, group images per station via the Workspace pane or Python.
  5. Run matchPhotos / alignCameras.

The reconstructed 3D model is built from the parallax between stations, with each station contributing extensive scene coverage from one viewpoint.

Caveats

  • Not the same as MultiplaneLayout. Camera stations group per-image-orientation under one optical centre; MultiplaneLayout groups per-instant captures across multiple sensors with fixed inter-sensor offsets. They solve different problems — see Declaring a fixed-geometry multi-camera rig in Python.
  • No automatic detection. Camera stations must be declared manually (GUI grouping or Python API). EXIF metadata does not carry station information.
  • Nodal-point calibration. The "nodal point" of a lens is the entrance pupil, not the front element or the sensor. Calibrating the tripod head's rotation axis to coincide with the entrance pupil takes practice; mis-calibrated nodal heads reintroduce parallax that the station model cannot handle.

References

  • Metashape Pro User Manual (2.3) § Camera station data (p. 39) — describes the feature and the "Set Group Type → Station" workflow.
  • Metashape Python API Reference (2.3.1): CameraGroup.Type.Station, Camera.group, Chunk.addCameraGroup.
  • Forum thread, Best practices for tank scanning — the nodal-head requirement (msg 17376).

See also