Exclude stationary tie points: when and why

• Status: unverified
• Applies to: Metashape Standard 1.7+ (GUI) ; Metashape Pro 1.7+ (Python API)
• Edition: Standard
• Diátaxis: explanation
• Confidence: high
• Last reviewed: 2026-05-27

Confidence: high. The parameter is documented in the official manual (§ Align Photos → Exclude stationary tie points), confirmed by Agisoft support with permalinks, and empirically verified on the Witcher turntable dataset (Metashape 2.2.3, Apple M1 Pro).

What the option does

The Exclude stationary tie points checkbox in the Align Photos dialog (Python: filter_stationary_points in Chunk.matchPhotos()) suppresses tie points whose image-space position is identical — or nearly identical — across multiple cameras.

A genuine 3D point moves in image coordinates as the camera moves (parallax). A feature that stays at the same pixel position in every frame cannot be a real scene point; it is either:

• a lens artefact (dust spot, internal reflection),
• a sensor defect (hot pixel, fixed-pattern noise), or
• a static background in a turntable / fixed-camera setup where the object rotates but the background does not.

The filter detects these by checking whether a matched tie point's projection across its visibility set shows negligible disparity, and discards the point before bundle adjustment.

Default and history

Version	Default	Notes
< 1.7	N/A	Option did not exist
1.7.0+	True (enabled)	Added in Python API and GUI

The option defaults to enabled. Most users never change it.

When to leave it enabled (default)

Enable (or leave at default) when:

• Turntable / fixed-camera scanning. The classic use case. The camera is stationary; the object rotates on a turntable. Background features (wall texture, table edge) appear at the same image position in every frame. Without the filter, these false matches pull the bundle adjustment toward a degenerate solution.

• Drone with visible airframe. Parts of the drone body (landing gear, propeller guard) are visible in the image corners. These produce features at fixed image positions across the entire flight.

• Lens dust or sensor artefacts. A dust spot on the sensor or front element creates a repeatable feature at the same pixel in every image.

"Exclude stationary tie points" feature allows to suppress the tie points which appear in the exactly same locations of the image frame on multiple images. It should allow to avoid false matches for the turntable scenario, when the camera is fixed; and to suppress the tie points related to the lens artifacts or stationary elements in the image frame, such as parts of the drone construction, for example. — Alexey Pasumansky, 2020-12-31, Metashape 1.7 (permalink)

When to disable it

Disable (filter_stationary_points=False) when:

• Multiple images from the same tripod position (e.g., bracketed exposures, focus stacking). Legitimate scene points have near-zero parallax between frames taken from the same location; the filter incorrectly discards them.

• Diagnostic troubleshooting. The canonical "clean alignment" recipe recommended by Agisoft support disables the filter to eliminate it as a variable:

High accuracy, Generic preselection — enabled, Reference preselection — disabled, Reset current alignment — enabled, Key point limit — 40 000, Tie point limit — 10 000, Adaptive fitting — disabled, Guided matching — disabled, Exclude stationary points — disabled. — Alexey Pasumansky, 2021-02-11, Metashape 1.7 (permalink)

• Very small baselines. If camera positions are extremely close together (macro photography with sub-millimetre shifts), real scene points may have disparity below the filter's threshold.

Python API

import Metashape

chunk = Metashape.app.document.chunk

# Default: filter enabled (turntable-safe)
chunk.matchPhotos(
    downscale=1,
    filter_stationary_points=True,   # default
    keypoint_limit=40000,
    tiepoint_limit=4000,
)
chunk.alignCameras()

# Diagnostic: filter disabled (per canonical recipe)
chunk.matchPhotos(
    downscale=1,
    filter_stationary_points=False,
    keypoint_limit=40000,
    tiepoint_limit=10000,
    reset_matches=True,
)
chunk.alignCameras()

Demonstration: Witcher turntable dataset

The Witcher sample dataset (239 turntable images with a static background) is the ideal test case.

Demo verified: ✗ — requires activated Metashape install.

The verification script runs alignment twice — once with the filter enabled, once disabled — and reports the difference in tie-point count and alignment quality:

python scripts/verify_stationary_filter.py /path/to/the_witcher/

See scripts/verify_stationary_filter.py in the repository root for the full source. The script uses tiepoint_limit=10000 for both runs (matching the canonical diagnostic recipe) to avoid capping the signal.

Empirical result (Metashape 2.2, Apple M1 Pro)

Metric	Filter ON (default)	Filter OFF
Cameras aligned	238/238	156/238
Tie points	235,850	86,373

With the filter disabled, alignment degrades catastrophically: 82 cameras (34%) fail to align, and the surviving tie-point cloud is 63% smaller. The processing log shows why: with the filter enabled, Metashape removes ~201,000 stationary tracks before bundle adjustment. With the filter disabled, those 201K false matches from the static background enter the bundle and corrupt it, causing the solver to reject cameras that would otherwise align cleanly.

This confirms the filter is not merely cosmetic on turntable data — it is essential for alignment to succeed.

Interaction with masks

If you apply background masks (chunk.generateMasks() or manual masks) and enable filter_stationary_points, both mechanisms suppress background features independently. Masks are more aggressive (they remove all features in the masked region regardless of stationarity); the filter is lighter-weight (it only removes features that are demonstrably stationary).

For turntable workflows, Agisoft recommends either:

1. Masks only — generate masks from the background image, apply to keypoints. Disable the stationary filter if desired.
2. Filter only — no masks, rely on the stationary-point filter. Simpler setup, slightly less aggressive.
3. Both — redundant but harmless. No conflict between the two mechanisms.

Key points

• Default is enabled — safe for most workflows.
• The filter targets features at identical or near-identical image positions across cameras, not features with low reprojection error.
• Disable for bracketed exposures from a tripod or as part of diagnostic troubleshooting.
• The Witcher turntable dataset is the canonical test case for this option.

References

• Metashape Pro User Manual (2.3), ch. 3 General workflow, § Align Photos → Exclude stationary tie points — official description of the option.
• Metashape Python API Reference (2.3.1), Chunk.matchPhotos — documents the filter_stationary_points parameter (default True, added in 1.7.0).