Dense Point Cloud, DSM, and Orthomosaic Generation

Dense Point Cloud, DSM, and Orthomosaic Generation

Why This Workflow Matters

After your UAS imagery is aligned and your Ground Control Points (GCPs) are imported, you have a sparse point cloud and camera positions—but no deliverable products yet. The three stages covered in this lab—dense matching, surface model building, and orthomosaic export—transform raw alignment data into the georeferenced files that engineers, surveyors, and planners actually use. Each stage builds on the one before it, so errors introduced early propagate forward. Working carefully and logging every setting is not busywork; it is how you prove your results are reliable.

Stage 1: Build Dense Cloud

Dense image matching compares overlapping pixels across many photos and reconstructs a dense three-dimensional point cloud—sometimes tens of millions of points for a modest survey area. In Agisoft Metashape the key parameter is Quality: Ultra High, High, Medium, Low, or Lowest. Higher quality uses full-resolution images and produces finer detail but demands far more RAM and processing time. For most survey-grade work, High is appropriate; Medium is acceptable for preliminary products or when computing resources are limited.

The Depth filtering setting controls how aggressively Metashape removes points that are statistically inconsistent with their neighbors. Mild filtering retains more points but may keep noise; Aggressive filtering is cleaner but can remove legitimate breakpoints on edges and roofs. For topographic surveys, Mild or Moderate filtering is typical.

Record your chosen quality and filtering mode in your processing log along with the elapsed time. Processing time varies widely with hardware—a dataset of 200 images at High quality might take 30 minutes on a workstation with a capable GPU or several hours without one.

Stage 2: Build DEM (Digital Elevation Model)

From the dense cloud, Metashape interpolates a raster elevation surface called a Digital Surface Model (DSM) when it includes vegetation and structures, or a Digital Terrain Model (DTM) when ground-level filtering has been applied. In Metashape the menu item is Build DEM; choose the dense cloud as the source.

Set the coordinate system to match your GCPs (typically a projected system such as a State Plane zone in US Survey Feet or meters, combined with a vertical datum). Set the resolution to at or better than your target Ground Sample Distance (GSD). If your flight was designed for a 3 cm GSD, exporting the DEM at 5 cm is acceptable; exporting at 1 m throws away information. Record the output resolution in your log.

Stage 3: Build Orthomosaic

An orthomosaic is a geometrically corrected, mosaicked image in which displacement caused by terrain relief and camera tilt has been removed using the DEM. Every pixel in a true orthomosaic maps to a consistent ground location at a known scale, making measurement possible directly on the image.

In Metashape, Build Orthomosaic uses the DEM you just produced. Select the same coordinate system and a pixel size at or near your design GSD. The blending mode (Mosaic is the default) controls how overlapping image regions are merged; Mosaic prioritizes nadir-facing pixels and generally produces the cleanest result for mapping.

GCP Import and Optimization

Before building dense products, you must import your GCP coordinates, mark targets on the photos, and run Optimize Cameras. Import the GCP coordinate file (CSV or text) using the Reference panel. For each GCP, Metashape prompts you to click the target center in multiple images—aim for at least three images per GCP; more is better. After marking, run Optimize Cameras.

Metashape reports two key accuracy indicators:

• Reprojection error (pixels): the average distance between where the software predicts a tie point or GCP should appear in each image and where it was actually marked. A value below one pixel is generally acceptable; significantly higher values suggest poor target marking or camera calibration issues.
• GCP residuals (ground units): the difference between your imported GCP coordinates and the model's computed positions. These residuals must fall within the accuracy tier you are claiming for your deliverable.

Document pre-optimization and post-optimization values in your processing log. If residuals remain large after optimization, re-examine your target markings before proceeding.

Export and File Management

Export three files:

1. Orthomosaic — GeoTIFF, full resolution, coordinate system matching GCPs.

2. DEM — GeoTIFF, matching resolution and coordinate system.

3. Processing report — PDF generated by Metashape (Export → Generate Report), which records camera positions, GCP residuals, overlap statistics, and software settings.

Follow the naming convention supplied by your instructor (typically: ProjectName_Date_ProductType, for example `SurveyA_20250610_Ortho.tif`). Store all files in the provided project folder structure and back up immediately to the external hard drive before closing Metashape.

Open both GeoTIFFs in ArcGIS Pro and compare them to a basemap. Roads, building corners, and other identifiable features should align visually. Significant offset indicates a coordinate system mismatch or a GCP import error—do not submit until alignment is confirmed.

Workstation Safety and Data Hygiene

Processing large datasets stresses hardware. Do not block workstation vents, and do not run other intensive applications simultaneously. Save your Metashape project frequently—dense cloud generation is not automatically saved and a crash means restarting that stage. Treat all source imagery as irreplaceable: never edit original files; always work from copies in your project folder. Back up before and after each major processing stage.