Quality Control: Checkpoints, Accuracy Reports, and ASPRS Standards

Why Quality Control Matters in Photogrammetry

A photogrammetric deliverable — an orthomosaic, a digital elevation model (DEM), or a point cloud — is only as trustworthy as the accuracy assessment that accompanies it. Clients use these products to make decisions about construction staking, cut-and-fill volumes, floodplain mapping, and property boundaries. An unverified product can send those decisions in the wrong direction. Quality control (QC) is the structured process of checking a finished dataset against independent measurements to confirm it meets a defined accuracy standard before delivery.

The industry benchmark for positional accuracy in the United States is the ASPRS Positional Accuracy Standards for Digital Geospatial Data, published by the American Society for Photogrammetry and Remote Sensing. These standards define accuracy classes for horizontal and vertical components and specify how to compute and report error statistics. Understanding and applying them is a core professional skill.

Ground Control Points vs. Checkpoints

Ground control points (GCPs) are surveyed targets that are fed into the photogrammetry processing software to georeference and refine the model. Because the software uses GCPs to build the solution, GCPs cannot objectively verify it — that would be like grading your own exam.

Checkpoints (also called independent check points, or ICPs) are surveyed targets that are deliberately withheld from the processing software. After the model is complete, you extract the model's predicted position at each checkpoint location and compare it to the independently measured GNSS position. The difference reveals the true error of the dataset.

A minimum of five well-distributed checkpoints is standard practice for small-area drone surveys. Checkpoints should be spread across the project area — corners, center, and mid-edges — not clustered together.

Computing RMSE

The standard error metric is Root Mean Square Error (RMSE). It is calculated separately for the horizontal components (X and Y, or easting and northing) and the vertical component (Z, or elevation).

For a single axis, the RMSE formula is:

```

RMSE = sqrt( (1/n) * Σ(measured_i - reference_i)² )

```

Where n is the number of checkpoints, measured_i is the dataset value at checkpoint i, and reference_i is the GNSS-surveyed value.

Horizontal accuracy is reported as a combined value called RMSEr (radial), derived from the X and Y components:

```

RMSEr = sqrt( RMSE_X² + RMSE_Y² )

```

The ASPRS standards use these RMSE values to assign an accuracy class. Each class has a maximum allowable RMSE for horizontal and vertical. Any individual checkpoint error that exceeds a defined multiple of the class threshold (typically three times the RMSE threshold) is flagged as a blunder and must be investigated.

ASPRS Accuracy Classes

The ASPRS table lists accuracy classes by their horizontal and vertical RMSE thresholds. Lower numbers mean higher accuracy (tighter tolerances). For example, a higher-accuracy class might permit only a few centimeters of RMSE, while a lower-accuracy class permits several decimeters. You determine the class your dataset achieves by finding the tightest class whose threshold your computed RMSE does not exceed. If your RMSE is worse than the loosest class, the product fails classification and the cause must be identified.

Always consult the current published ASPRS standard document — thresholds are specific and must not be recalled from memory for reporting purposes.

Extracting Values in ArcGIS Pro

In ArcGIS Pro, use the Extract Values to Points geoprocessing tool (or the Sample tool for rasters) to pull the DEM elevation value at each checkpoint location. For horizontal comparison, use the orthomosaic and visually confirm checkpoint center locations with the Identify tool or by reading coordinates from the checkpoint target layer. Record all extracted values and reference values in the QC spreadsheet before performing any calculations.

The QC Memo

The deliverable to a client is not just a dataset — it is a documented accuracy report. A professional QC memo must include:

1. Mission parameters — flight date, altitude, overlap, sensor used.

2. Processing settings — software, image alignment quality, GCP count and distribution.

3. Checkpoint statistics — individual errors, RMSE_X, RMSE_Y, RMSEr, RMSE_Z.

4. ASPRS accuracy class achieved — stated clearly.

5. Flags and recommendations — any checkpoint exceeding the blunder threshold, and a recommendation to re-fly or reprocess if warranted.

Write the memo in plain language. Avoid jargon that a non-technical client cannot understand.

Safety and Data Integrity

Data QC carries its own safety dimension. Inaccurate elevation data used for drainage design, road grading, or utility mapping can lead to engineering failures and harm to people and property. Never alter or delete checkpoint measurements to make statistics look better. Report results honestly, even if the dataset fails a desired accuracy class. If a blunder exists, investigate the cause — it may indicate a problem with a GCP, a processing error, or a GNSS measurement that needs to be re-collected in the field. Protecting the integrity of the data is a professional and ethical obligation.