Skip to main content

PCM (Point Cloud Merger)

ADDON
📖4 min read
How to use Point Cloud Merger in Blender - Part 1 by Olli Huttunen
1 / 2

Introduction​

PCM (Point Cloud Merger) is a paid Blender addon by ToppiNappi (€26) that merges multiple COLMAP scans into one clean dataset. Scan the same environment from different areas or platforms: ground level and aerial, LiDAR and 360° camera. PCM imports, aligns, and merges the reconstructions visually in the viewport, then exports a single COLMAP package ready for Gaussian Splatting training.

Key Takeaways​

Merge multi-pass scans

Combine LiDAR, ground, and aerial COLMAP scans into one clean training dataset.

RMS-quantified alignment

Point-pair alignment with an error readout. Use 3 pairs minimum, 5+ for a better fit.

Trainer-ready export

Standard COLMAP folder out. Drag into Postshot or any 3DGS trainer and go.

Installation​

  1. Purchase from the Gumroad listing (€26).
  2. Drag the downloaded .zip onto the Blender viewport and confirm the install prompt.
  3. Press N in the viewport to open the side panel. Work from the Point Cloud Merger tab.

Usage​

The workflow spine is import → level → align → export, from the Point Cloud Merger panel (press N in the viewport to open it).

Import the scans· import

Browse to each COLMAP dataset directory (where images/, cameras, and points3D live; .bin or .txt both work) and press Import. Use Orientation to rotate a scan 90° at a time until it sits naturally, and rename datasets in the Reconstructions list.

Level a tilted scan (optional)· leveling

Select a leaning cloud, press Mark Level Point, pick three points on a flat surface (e.g. the ground), then press Straighten to XY Plane to rotate the whole cloud onto that plane.

Align the second scan· align

Select the reference cloud, mark matching source points, find the same points on the target cloud (3 pairs minimum, 5–6 better, spread across the scene at different heights), then press Align Target to Source. The panel reports an RMS error.

Export the merged package· export

With both clouds visible, press Export Merged COLMAP Package, confirm the summary totals, pick an output folder, and wait for the export notification. The output is standard COLMAP layout (images/ + sparse/), ready to feed a 3DGS trainer.

In an Investigation​

PCM solves a specific multi-scan problem: combining a precise but limited pass (LiDAR, ground) with a wider but looser pass (360° camera, drone, pole) into one training dataset. The visual, point-pair alignment keeps the merge inspectable. The RMS error gives you a quantified quality signal to cite in your methodology.

What to document when using PCM

In Your Method Section​

State the source datasets merged (platform, capture method), the PCM version, and the alignment result (pair count, RMS error).

In Your Decision Log​

  • Which scan was chosen as reference and why.
  • Number and placement of alignment point pairs.
  • Leveling choices (which plane defined as ground).
  • RMS error of the final fit, and any manual nudges after it.
  • Date performed.

Verification​

Re-import the exported COLMAP package into a fresh Blender file or trainer and confirm orientation, camera placement, and point alignment before training.

Limitations​

  • Requires two or more COLMAP reconstructions with overlapping coverage.
  • Paid addon (€26, Gumroad).
  • Alignment is only as good as your point pairs; residual offset may need manual correction.

Resources​