Extract a skeleton

Pull the branching topology out of a woody scan: a graph of nodes (junctions and tips) connected by edges (branch segments), with per-branch order (Strahler number) and length.

Inputs

You need:

• Exactly one selected point cloud of a woody plant (typically TLS data, ground removed). Best results with even coverage and intensity > ~1k points/dm³ on branches.

Skeleton extraction runs on point clouds only — a mesh is not a valid input, and the tool stays greyed out unless a single cloud is selected.

Leaves can be present but degrade results — if you have a way to filter the cloud to woody points only (intensity threshold, classification field), do so first.

Run extraction

1. Select the cloud in the Scene panel.
2. Click Extract Skeleton (the DNA-helix icon in the Tools › Reconstruction group).
3. The Skeleton Extraction (BFS Graph) panel opens on the right. There is one algorithm — a BFS graph method — so there's no method to choose.

4. Set the options you need:

   • Remove outlier points — checkbox; drops isolated noise before building the graph.
   • Smooth skeleton (Laplace) — checkbox; Laplacian smoothing of the resulting centerline.
   • Search Radius — slider, 0–0.2 m. Left at 0 it reads Auto (based on density) and picks a radius from the cloud itself.
   • Min Points/Block — slider, 1–50 (default 5). Raise it to suppress spurs in noisy regions.

5. Advanced Options (collapsed by default) adds Root Threshold, Quantization Levels, Nonlinear quantization, Proportion filter, and Smoothing Iterations.

6. Click Extract Skeleton. While it runs the button shows a spinner and a Cancel button appears beside it — Cancel kills the computation and adds nothing.

A new entry appears in the Skeletons list, showing its total length and point count (e.g. 12.84m · 512 pts).

Visualize branch order

Skeletons draw in a single flat color (amber) by default. Tick Color by branch order in the Skeletons panel to color by branch order instead: an 8-step ramp indexed from the tips — order 1 red, then orange, yellow, green, cyan, blue, violet, and order 8+ (the trunk) pink. Orders past 8 clamp to pink.

The same panel has Show as cylinders, which draws the edges as tubes instead of lines, with a Tube Radius slider.

Get measurements out

The skeleton's row reports its total length and point count.

For anything more detailed, export it:

• .json — every node (position + branchOrder) and every edge as a [from, to] index pair, plus totalLength, nodeCount, edgeCount, and maxBranchOrder. This is the only format that imports back into Phytograph. See Skeleton JSON shape.
• .obj / .ply — geometry for Blender or MeshLab.

Per-edge lengths aren't written out; derive them from the node positions.

Common problems

"My skeleton has spurs everywhere." The input cloud is noisy, or the graph is being built too finely. Raise Min Points/Block, tick Remove outlier points, and consider separating wood from leaf first so only woody points feed the extraction.

"The skeleton is missing whole branches." Coverage in those regions is too sparse to connect. Solutions:

• Raise Search Radius (or leave it at Auto) so the graph bridges wider gaps
• Combine multiple scan positions via Stitch before extracting
• Lower Min Points/Block, which may be discarding thin branch tips

"The trunk is in two pieces." Usually a connectivity gap near the ground. Make sure ground points are removed but the trunk base is intact — overcropping the bottom can disconnect the root of the skeleton.

What's next

• Compare to a generated plant model of the same species — useful for validating both your scan and the model.
• Export the skeleton as .json for analysis in Python or R.