This article explains how to generate a DEM for use in orthophoto reconstruction. There are three possible DEM options, ranging from lowest to highest quality.

1.Prepare the Point Cloud

After finishing this step go to the next section to export the DEM.

1.1. Quick DSM

Quick to generate, irregular features, like trees, that tend to get distorted and edges around buildings may be distorted.

a. Remove high and low noise

More details on the high quality DEM section.

After finishing this step go to section 2 to export the DEM.

1.2. Quality DEM

Quality DEM, all the features will be perfectly model, edges around buildings may be a bit distorted.

a. Remove high and low noise

More details on the high quality DEM section.

b. Run ground classification

More details on the high quality DEM section

c. Classify roofs and bridge decks

More details on the high quality DEM section

After finishing this step go to section 2 to export the DEM.

1.3. High quality DEM

High Quality DEM, all the features will be perfectly model.

a. Remove high and low noise

Clean the dataset by removing obvious high and low noise. This is often best done using one or more of the following:

• Clip Range in Geocode LiDAR
• Low/Isolated PCT
• Basic Filter PCT with elevation clipping

See Noise Removal Options for additional guidance.

b. Run ground classification

Run Adaptive TIN Ground PCT using settings appropriate for a clean, generalized ground surface. The surface does not need to be perfect for orthomosaic generation, but it should reasonably represent the terrain. In areas with features such as retaining walls, breaklines may be needed to preserve those edges properly.
 

c. Optional: Clean up major ground blunders

Optionally clean obvious errors in the ground surface (Class 2) using:

• the manual classification tools on the Classification tab and Profile toolbar
• Ground Cleanup Filter PCT

d. Classify roofs and bridge decks

For a better-looking orthomosaic, manually classify:

• building roofs to Class 6
• bridge decks to Class 17

e. Optional: Improve building and bridge edges (BIT License required)

To create the cleanest edge lines along buildings and bridges, polygonize the building and bridge deck classes and assign them to both ground and upper-surface elevations. This process creates two polygon sets:

• buffered footprints
• headprints

Follow these steps:

1. Use Point Group Tracing and Squaring PCT with parameters appropriate for your dataset to polygonize the desired non-ground features. These are typically roofs and bridge decks, but may include any features intended for a True Ortho workflow.
   For high-density TrueView data, first thin the dataset to approximately 8 points/m² using Classify by Statistics PCT.
2. Use Conflate PCT to purely drape a copy of each non-ground feature onto the ground class, creating the footprints. This keeps the feature edge clean against the existing ground surface.
3. Use Conflate PCT again to conflate a copy of each footprint polygon to the closest elevation of the building or bridge deck points contained within it. This creates the headprints.
4. Buffer the footprint polygons slightly, for example 2.5 cm, using the Buffer Geometry tool on the Feature Edit tab. This is necessary because a TIN cannot contain two points at the same planimetric location with different elevations.
5. Use Classify by Feature PCT to mark any ground points found within the buffered footprint polygons so they are excluded from the final DEM.

6. If you complete this workflow and generate both:

   • buffered footprint polygons (buffered, ground-conflated polygons), and
   • headprint polygons (closest-elevation conflated polygons),

   then you should skip the Classify by Statistics PCT in the next step, since there will no longer be conflicting surfaces in the same cell.
   In addition, enable Breakline Enforcement during DEM export and use the polygons created above to better preserve feature edges.

f. Run Classify by Statistics PCT

For best results, perform the initial classification steps above before exporting the DEM. Then run Classify by Statistics PCT using the following recommended settings. This is especially helpful in areas where two surfaces occupy the same location, such as where roof overhangs allow LiDAR to see the ground below.

Recommended settings:

1. Feature Geometry – Tool Geometry
2. Units – Set to project units
3. Input LAS Layer – Active LAS Layer
4. Source Points – Classes 2, 6, and 17; flags ignored
5. Cell Size – 0.05 m
   Gently rolling terrain does not require as high a DEM resolution as terrain with sharper features.
6. Samples – Max
7. Destination Class – Flags Only → Synthetic → Set
8. Quartile Classification – Leave blank
9. Generate Cell Output Shapefile – Leave unchecked
10. Run by project

After finishing this step go to the next section to export the DEM.

2. Export the DEM

Use the Export Wizard to quickly generate a DEM from the LiDAR data. This approach is typically much faster than having Metashape create the DEM from images.

ii. Recommended Export LiDAR Data DEM settings

1. Source Points

Choose one of the following:

• If Classify by Statistics PCT was performed in the previous step, use Synthetic Flag Set
• If the optional non-ground polygons were created and Classify by Statistics PCT was skipped, use Classes 2, 6, and 17, ignoring the synthetic flag
• For a quick ortho, you may also use all points

2. Export Type – Surface
3. Surface Method – Triangulation
4. Pixel Size – 0.05 m to 0.25 m
Gently rolling terrain does not need as high a DEM resolution as terrain with sharper features.
5. Surface Attribute to Export – Elevation
6. Export Format – GeoTIFF
7. Raster Information → Pixel Size – 0.05 m to 0.25 m

3. Use the DEM for the Orthophoto generation

Learn how to use Agisoft Ortho Mapping