CHM¶

A Digital Elevation Model (DEM) is a bare-earth terrain surface — a regular grid of ground elevations. Phytograph builds one from a cloud's ground points, interpolating their elevation onto a grid and reconstructing a heightmap surface. It's the natural follow-on to ground segmentation: segment the ground first, and the DEM is built from the classified ground returns rather than the whole cloud.

The same tool also builds two related surface products. Tick any combination in the Surfaces list at the top of the panel — one run generates every checked product, each as its own mesh (so you can produce a DTM, DSM and CHM in a single click):

Terrain (DTM) — the bare-earth ground surface (the classic DEM; the default).
Surface (DSM) — the first-return / top-of-canopy surface: the highest return in each cell. It does not need ground classification.
Canopy height (CHM) — the canopy height model, DSM − DTM: vegetation height above the bare earth. Phytograph grids the ground (DTM) and the first-return surface (DSM) on one aligned grid and subtracts them, so every cell reads the height of the canopy above the ground beneath it (never negative). See Terrain surfaces: DTM, DSM, CHM for the concepts.

Terrain layers (density, intensity, hillshade, slope, aspect)¶

A generated Terrain (DTM) doesn't just carry elevation — it comes with a set of scalar layers you colour the same surface by and export individually. No checkboxes and no extra runs: the layers are computed automatically with the DTM. Expand the DTM in the Meshes panel and pick a band from Color by:

Elevation — the ground height (the default).
Point density — number of points per cell (a coverage/density map).
Return density — number of laser pulses (first returns) per cell.
Intensity — mean return intensity per cell (present when the cloud carries an intensity field).
Hillshade / Slope / Aspect — shaded relief (fixed sun, 315°/45°), steepness, and facing direction, derived from the elevation grid (GIS-standard).

The colorbar re-captions to the selected band, and Export raster (below) lets you write any subset of these layers to .asc / GeoTIFF.

Generate¶

Select a single point cloud.
(Recommended, for DTM/CHM) Run Segment Ground first. The DEM tool is ground-class aware: if the cloud carries a ground classification, only the ground points are gridded, giving a true bare-earth model. If it doesn't, the panel shows a notice and the ground is auto-detected with the Cloth Simulation Filter (its cloth settings are scaled to the cloud's size, so a field- or ALS-scale tile is handled as efficiently as a close-range scan). Running Segment Ground first just gives you control over that step (and lets you reuse the classification elsewhere).
Click Generate DEM (the mountain icon in the tool column), or open the command palette and choose Generate DEM (searching "DSM" or "CHM" also finds it). The Generate surfaces panel opens.
Tick the Surfaces you want (Terrain / Surface / Canopy height) — any combination, generated together in one run. DTM and CHM use the ground classification; DSM does not.
Adjust the parameters if needed (hover the ? beside any for a quick explanation):
- Cell size (m) — the horizontal resolution of the grid. Smaller resolves finer terrain but runs slower and leaves more gaps where the ground is sparsely sampled; larger is coarser and smoother. It's seeded from the cloud's extent each time the panel opens — a few centimetres for a close-range scan, larger for a field-scale tile.
- Interpolation — how elevation is filled between ground points. TIN (linear) builds a triangulated surface through the ground returns; it's the most faithful and the default. IDW smooths across neighbours. Nearest snaps each cell to the closest ground point (blocky but gap-free).
- Fill data gaps — off by default, so cells with no ground return stay empty (an honest DTM of only what was measured). Turn it on to fill any cell that has returns but no ground — e.g. under dense canopy, where few pulses reach the ground and the bare-earth model would otherwise be full of holes — with the nearest measured ground elevation, so the DTM covers the whole scanned footprint. Either way the surface is clipped to that footprint: it never extrapolates past the scanned area (no fabricated corners beyond a rotated survey tile).
- Compute height above ground (Terrain/DTM only) — also subtract the DEM from each point to add a height_above_ground scalar to the cloud (a per-point canopy-height precursor). Off by default. For a rasterised canopy height model, use the Canopy height (CHM) surface instead.
Click the run button. Its label follows your selection: Generate DEM, Generate DSM, or Generate CHM for a single surface, or Generate 2 surfaces / Generate 3 surfaces for several. With nothing ticked it reads Select a surface and stays disabled. While it runs, the button shows a spinner (with the gridding progress) and a Cancel button appears beside it — click Cancel to stop a long or stuck run immediately (the computation is killed and nothing is added).

Watch the grid estimate

The panel shows a live Estimated grid: nx × ny cells readout. Past 4,000,000 cells it turns red, adds "— too fine; increase cell size", and disables the run button. If you can't start a run, that's usually why: raise the cell size.

For a DTM, robustness against residual non-ground points (low vegetation that slipped into the ground class, stray low noise) comes from representing each grid cell by a low percentile of the heights that fall in it before interpolation, so a few high outliers can't tent the terrain. A DSM does the opposite — a high per-cell percentile — so each cell tracks the top of the canopy rather than a low outlier. A CHM subtracts the two on an aligned grid and applies a first-pass pit-fill so isolated within-canopy dips don't read as holes.

Inspect and use the result¶

The surface appears as a new surface mesh in the scene, named … DEM, … DSM, or … CHM, coloured by elevation (for a CHM, that gradient reads as canopy height). Like any mesh you can change its colour mode, opacity, and transform from the Meshes panel, and hide or delete it. A DTM additionally lets you switch its Color by dropdown between the terrain layers — the same mesh recolours by density, intensity, hillshade, and so on.

If you built a DTM with Compute height above ground ticked, the source cloud gains a continuous height_above_ground attribute and recolours by it (a gradient with a numeric colorbar). Switch back to it any time from the Color by picker.

A DEM also drives terrain following for leaf-area density: expand a voxel grid's row in the Meshes panel and click Snap to ground to displace the grid so it rides this surface — each column then measures the same height above sloping ground, and the LAD inversion uses the displaced grid you see. See Terrain following.

Export¶

GIS raster — expand the surface (DEM/DSM/CHM) in the Meshes panel and use Export raster: GeoTIFF (.tif) or ESRI ASCII grid (.asc), for use in QGIS / ArcGIS. (Note: .asc here is a raster grid, distinct from the .asc point-cloud import format.) For a DTM, tick which layers to write (elevation / density / intensity / hillshade / slope / aspect) — the band you're currently viewing is pre-checked; picking several writes one file per band into a chosen folder (<cloud>_<layer>.tif). DSM/CHM export their single grid. The GeoTIFF is georeferenced when the source cloud's CRS is known (e.g. a UTM LAS).
Mesh — to export the surface geometry itself, select the DEM and open Export (File → Export) for OBJ / PLY / STL, like any mesh.

Georeferencing

The raster is written in the cloud's own coordinates. For a cloud imported in a projected CRS (e.g. UTM easting/northing), the GeoTIFF carries the correct pixel size and origin so it lines up in GIS; assign the CRS on load if it isn't embedded.

Large clouds

Clouds imported from disk stream as an octree. DEM generation reads the in-memory session at full resolution and the per-cell pre-binning bounds the work to the grid size, so it stays fast even on multi-million-point tiles.