Accuracy and Quality Control

Why It Matters

A client asks: “How accurate is this map?” If you can’t answer with a specific number and explain how you determined it, you’re not delivering professional mapping services. You’re delivering guesswork.

Accuracy quantification separates professional mappers from hobbyists. Construction decisions, legal boundaries, and million-dollar earthwork contracts depend on knowing that your data is reliable. This lesson covers how accuracy is measured, what affects it, and how to verify it.

Understanding Accuracy Metrics

Root Mean Square Error (RMSE)

RMSE is the standard metric for mapping accuracy. It measures the average magnitude of error between known points and their corresponding positions in your processed model.

RMSE is calculated for three components:

• RMSEr: horizontal (radial) error in X and Y
• RMSEz: vertical error in Z (elevation)
• RMSEt: total 3D error

Lower RMSE means higher accuracy. A mapping project with RMSEr of 2 cm means the average horizontal error between any point in the model and its true position is 2 cm.

Accuracy Targets

Application	Horizontal RMSE	Vertical RMSE
Visual/Marketing	30+ cm	50+ cm
General construction	5-10 cm	10-15 cm
Earthwork volumes	3-5 cm	5-10 cm
Topographic survey	2-3 cm	3-5 cm
Engineering design	1-2 cm	2-3 cm

These targets assume proper GCPs and processing. Without GCPs, you’re in the “visual” category regardless of your equipment.

Common Error Sources

Data Collection Errors

1. Motion blur: the most common accuracy killer. Photos with even slight blur don’t align properly, degrading the entire model. Use fast shutter speeds (1/1000+) and fly slowly.

2. Incorrect focus: out-of-focus photos fail feature matching. Set manual focus at infinity and verify before each flight.

3. Insufficient overlap: below 60% overlap, the software can’t reliably match features. Always use 70%+ side overlap and 75%+ front overlap.

4. Inconsistent altitude: large altitude changes within the mapping area create varying GSD, complicating processing.

5. Moving objects: vehicles, people, and animals create “ghost” features that confuse alignment.

Environmental Errors

6. Vegetation: dense grass, crops, and tree canopy obscure the ground surface. The DSM measures the top of vegetation, not the terrain beneath.

7. Water and reflective surfaces: photogrammetry can’t match features on water, glass, or highly reflective surfaces.

8. Shadows: moving cloud shadows change between photos, creating inconsistent feature matching.

9. Wind: strong wind causes the drone to pitch and roll, changing the camera angle and creating inconsistent capture geometry.

Processing Errors

10. Incorrect GCP marking: clicking even a few pixels off from the GCP center introduces systematic error.

11. Wrong coordinate system: a mismatch between your GCP survey and processing software shifts or warps the entire model.

12. Inappropriate settings: running processing at “Low” quality for speed produces less accurate results.

Validation Methods

Check Points

The most reliable validation method. Before processing, set aside 1-3 surveyed GCPs as check points. These points are NOT included in processing. The software doesn’t use them to align the model.

After processing, measure where these check points appear in your model and compare to their surveyed coordinates. The difference is your actual achieved accuracy.

Report Your Findings

A professional mapping report includes:

• RMSE values (X, Y, Z)
• Check point error table: each check point’s expected vs. measured position
• Number of GCPs and check points used
• GSD (resolution)
• Coordinate system and datum
• Software and processing settings
• Date and conditions of data collection

Visual Quality Checks

Beyond numerical accuracy, visually inspect your outputs:

• Orthomosaic: are there visible seams? Any blurred or misaligned areas? Do straight features (roads, buildings) appear straight?
• DEM: any unexpected spikes or holes? Does the terrain look realistic?
• 3D model: any “melted” areas where features didn’t reconstruct properly?
• Point cloud: are there areas with unusually low point density?

When to Re-Fly

Re-fly the site if:

• The orthomosaic has visible alignment errors
• RMSE exceeds your project’s accuracy requirement
• Large areas are missing or poorly reconstructed
• Photos show consistent blur or focus problems

Never rely solely on the software's reported accuracy. The software optimizes to minimize error at GCPs, which makes its internal accuracy metrics optimistic. Independent check points give the true picture. If you provide accuracy guarantees for mapping data used in construction or engineering decisions, those numbers have legal weight. Be conservative in your claims. Report what you measured, not what you expected.

Quick Check

Q: What does RMSE measure? A: Root Mean Square Error, the average magnitude of error between known ground positions and their corresponding positions in the processed model.

Q: What is a check point and how is it different from a GCP? A: A check point is a surveyed point NOT used in processing. After processing, its model position is compared to its known position to measure actual achieved accuracy.

Q: What is the most common cause of mapping inaccuracy? A: Motion blur. Photos with even slight blur degrade feature matching and alignment throughout the entire model.

What’s Next?

Final lesson: delivering results to clients in formats they can actually use, from Google Earth overlays to GIS integration.