How do Ground Control Points (GCPs) improve accuracy in drone mapping?

Ground control points (GCPs) improve drone mapping accuracy by tying the map to surveyed positions on the ground. Markers with precisely measured coordinates appear in the photos. Processing software then uses them to shift, scale and level the whole model. As a result, outputs can line up with real-world coordinates to within centimeters rather than meters.
Key takeaways
- Standard satellite positioning is accurate to meters, not centimeters. According to GPS.gov, GPS-enabled smartphones are typically accurate to within 4.9 m (16 ft) under open sky.
- One study covered more than 1,200 hectares. There, horizontal accuracy close to one pixel on the ground needed about 2.5 to 3 GCPs per 100 photos. Below 1 per 100, it fell away quickly.
- Placement matters as much as count. GCPs around the edge reduce horizontal error. About 0.5 to 1 GCP per hectare inside the site reduces height error.
- Checkpoints are surveyed points kept out of processing to test the result. The 2023 ASPRS accuracy standard asks for at least 30 for a fully compliant assessment.
- RTK and PPK drones need fewer GCPs. Even so, at least one is still advised, because direct georeferencing can leave errors in height.
What are ground control points (GCPs)?
Ground Control Points are specific locations on the ground that have been accurately surveyed and marked. These known reference points are used during drone data processing to help align outputs like orthomosaics or digital elevation models (DEMs) with real-world coordinates. In other words, they anchor the whole map to the right place on Earth.
Drone maps are a powerful way to understand what's happening on the ground, whether you're measuring stockpiles, checking site progress, or monitoring environmental change. But for those maps to be useful, they need to be more than just sharp images. They need to accurately reflect real-world positions. That's where Ground Control Points (GCPs) come in. GCPs are one of the most reliable ways to improve the positional accuracy of drone outputs.
People place physical markers, known as Ground Control Points, on the ground to define the location and orientation of an aerial image. These markers enable accurate georeferencing of the images, allowing precise measurements and analysis of the captured data. In photogrammetry, this is called indirect georeferencing. The model is tied to the coordinate system by points seen in the photos, rather than by logged camera positions alone.
How are GCPs placed and surveyed?
GCPs are usually flat markers, like checkerboard tiles, crosses, or painted targets, placed on the ground before a drone flight. They're designed to be clearly visible from above. Each one is also surveyed with high-precision equipment. Processing software then matches its known coordinates to where it appears in the photos.
Here's how the process typically works:
- Markers are laid out across the site, often on open ground or hard surfaces. These can be temporary (like portable tiles) or permanent (like painted symbols or concrete pegs).
- Each marker is surveyed using high-precision positioning equipment, giving an accurate latitude, longitude, and elevation, often within a few centimeters.
- The drone captures images, and the GCPs appear in the photos.
- During processing, the known coordinates are matched with their visual locations to georeference the entire dataset.
Depending on the project, GCPs might be placed by a surveyor, engineer, or a trained drone operator. After the flight, temporary markers are usually removed, although permanent GCPs are sometimes used for long-term monitoring or recurring surveys.
Temporary or permanent GCPs?
Temporary targets suit one-off surveys, because they are laid out and collected on the day. On a quarry or earthworks site surveyed every week or month, however, permanent marks save setting out targets each time. They also keep every survey tied to the same control. That matters as soon as you compare one survey with the next.
Permanent marks do need looking after, though. A mark that has been bumped or buried quietly shifts every map processed against it. Our guide to managing survey control covers how to store, re-survey and retire those points.
What does accuracy mean in drone mapping?
In drone mapping, there are two main types of accuracy to be aware of. Relative accuracy is how well features line up with each other within the dataset. Absolute accuracy is how well the whole dataset aligns with real-world coordinates. GCPs mainly improve absolute accuracy, and they also help keep heights consistent across a site.
- Relative accuracy refers to how accurate features are in relation to each other *within* the dataset. For example, the distance between two buildings or the slope of a road might be precise and consistent, even if the entire map is slightly offset.
- Absolute accuracy refers to how well the entire dataset aligns with real-world coordinates. In other words, are your features in the correct position on the Earth when compared to other geospatial data or surveyed ground points?
It's possible to have high relative accuracy without GCPs. Many drones, even with standard GNSS, can produce internally consistent outputs. That means distances, shapes, and volumes within the map can still be useful for certain tasks.
But without GCPs (or RTK/PPK corrections), those same outputs might not be correctly georeferenced, meaning they could be a few meters off from their actual position. If you're overlaying drone data with other spatial layers, tracking change over time, or submitting outputs for compliance, that global alignment matters.
Control is only one part of accuracy, though. Overlap, flight height and camera quality all play a part too. Our guide to drone mapping covers those from the start.
Why isn't a drone's own GNSS accurate enough?
Most drones are equipped with standard GNSS receivers, which tell the drone where it is when each image is captured. However, standard GNSS can have positional errors ranging from 1 to 5 meters due to things like satellite geometry, atmospheric interference, or terrain. For high-precision work, that's a wide margin of error.
GNSS stands for Global Navigation Satellite System, and includes satellite positioning systems like GPS (US), GLONASS (Russia), Galileo (EU), and BeiDou (China). According to GPS.gov, GPS-enabled smartphones are typically accurate to within 4.9 m (16 ft) under open sky. Accuracy also worsens near buildings, bridges, and trees. High-end receivers with dual-frequency signals or corrections, however, can reach a few centimeters in real time.
By contrast, using GCPs allows you to correct and align your outputs with known ground positions, improving accuracy and ensuring your results are georeferenced correctly to your chosen system. (Not sure how coordinate systems work? Here's a simple guide to coordinate reference systems.)
How much do GCPs improve accuracy?
GCPs typically move a drone map from meter-level to centimeter-level absolute accuracy. The size of the gain depends on how many GCPs are used, how they are spread and the image resolution.
| Without GCPs | With GCPs | |
|---|---|---|
| Map alignment | Visually aligned, but may drift from true coordinates | Anchored to real-world coordinate system |
| Elevation data | May be skewed or inaccurate | Tied to surveyed ground levels, within the accuracy of the control survey |
| Relative accuracy | Usually still consistent across the dataset | Maintains internal consistency and external alignment |
| Typical absolute error | A few meters with standard GNSS | Centimeters, often one to two times the GSD |
Image resolution sets the limit. Ground sample distance (GSD) is the distance on the ground between the centers of two neighboring pixels. Sanz-Ablanedo and colleagues tested 3,465 combinations of control points over a site of more than 1,200 hectares. According to their study, the GSD of a project relates directly to the best accuracy it can reach. So no amount of ground control makes a map more accurate than its pixels allow.
Small errors also add up quickly in a surface. On stockpiles and earthworks, a few centimeters of height error over a large area means a difference in cubic meters. Our article on GCPs in volumetric reporting explains why.
How many ground control points do you need?
There is no single number, but research gives useful ranges. One 2018 study found horizontal accuracy close to one GSD with about 2.5 to 3 GCPs per 100 photos. Below one per 100, accuracy degraded quickly. Another recommends GCPs around the site's edge plus about 0.5 to 1 per hectare inside.
In the Sanz-Ablanedo study, more than 3 GCPs per 100 photos did not meaningfully improve horizontal accuracy. Vertical accuracy, however, kept improving up to 4 GCPs per 100 photos, reaching about 1.5 times the GSD. That matches a classic rule in photogrammetry: heights need more control than positions. With very sparse control, horizontal error rose above 4 times the GSD and vertical error above 5 times. The authors note that their fit is specific to that site, so treat it as a guide.
In acres, 0.5 to 1 GCP per hectare is roughly one GCP every 2.5 to 5 acres. That density applies inside the boundary, in addition to the GCPs around the edge.
Small sites still need a minimum. In principle, three well-spread points are enough to fix a model's position, scale and rotation. With only three, though, there is no spare to show when one is wrong. That is why a few more are used in practice.
Where should GCPs be placed?
Placement matters as much as count. According to Martínez-Carricondo and colleagues (2018), edge and stratified layouts gave the best results of the distributions they tested. GCPs around the edge reduced horizontal (planimetric) error, while an even spread inside the site reduced height error. Combining the two minimized the total error.
In practice, that means:
- Around the edge: Near the corners and along the boundary, so the model is never stretched beyond its control.
- Through the middle: An even spread inside the site, with more where the ground rises or falls.
- On stable, open ground: Away from trees, buildings and anything that moves, so each target is clear in many photos.
- Inside the flight area: Every GCP should appear in several overlapping images, not just one or two at the edge.
Placement and flight design go together, so it helps to plan both at once. Our guide to drone flight planning covers overlap, altitude and coverage.
What is the difference between GCPs, checkpoints and levels?
Ground control points and checkpoints are both surveyed points on the ground, but they have different roles. GCPs are used during processing to fix the map in place. Checkpoints are kept out of processing and used afterwards to test how accurate the map is. Levels measure height differences precisely, so they give control points accurate heights.
GCPs are physical markers placed on the ground at known locations with precisely surveyed coordinates. They are used during the georeferencing process to align the drone imagery with the real-world coordinates of the GCPs. GCPs are typically placed throughout the area being surveyed to provide a network of reference points that can improve the accuracy of the georeferencing.

What are checkpoints in drone mapping?
Checkpoints are established at known locations on the ground that are separate from the GCPs. They are used to validate the accuracy of the georeferencing after the processing of the imagery is complete. Checkpoints are typically surveyed using high-precision equipment to determine their precise location. Their coordinates are then compared to the coordinates of the same locations in the georeferenced imagery. The difference between the two is used to calculate the accuracy of the georeferencing.
Checkpoints can be useful for several reasons:
- Quality control: Checkpoints can be used to ensure that the georeferencing is accurate and meets the required level of precision for the intended use of the imagery.
- Error detection and correction: Checkpoints can identify any errors or inaccuracies in the georeferencing process, which can be corrected to improve the overall accuracy of the imagery.
- Confidence assessment: Checkpoints can provide a measure of confidence in the accuracy of the georeferencing, which can be useful for decision-making.
The ASPRS Positional Accuracy Standards for Digital Geospatial Data (Edition 2, 2023) set the bar here. According to the standard, checkpoints should be an independent set of points that were not used in processing. They should also be at least two times more accurate than the product being tested. A fully compliant accuracy assessment needs at least 30 checkpoints, and accuracy is reported as root mean square error (RMSE).
The independence matters. Sanz-Ablanedo and colleagues found that accuracy is overestimated when it is measured only on control points. That is because the model has already been fitted to them, so independent checkpoints give the honest figure. On a small site, 30 checkpoints may not be practical. Even a handful kept out of processing, however, will catch a map that has shifted.
What are levels in aerial imagery?
Levels, also known as surveying levels or optical levels, are used to measure height differences or elevations in a given area. A level is an instrument, optical or digital, that is read against a graduated staff. Starting from a benchmark (a mark of known height), a surveyor carries heights from point to point. This method is called differential leveling.
It is the method national height datums were built on. For instance, according to Geoscience Australia, the Australian Height Datum was computed from 97,320 km of primary leveling. It was adopted in May 1971.
One of the main benefits of using levels is that they can provide highly accurate measurements of height differences or elevations in a given area. This can be particularly useful for applications such as surveying, construction, and infrastructure planning, where precise elevation data is needed.
In drone mapping, a level never touches the drone. Instead, its job is to give GCPs and checkpoints accurate heights. Another advantage of using levels is that they can be used to establish GCPs with high precision. Height is usually the weakest part of a drone survey, so better ground heights carry through to the elevation model.
Are GCPs, levels or checkpoints geotagging technologies?
Ground Control Points (GCPs), levels, and checkpoints are not typically considered geotagging technologies. Geotagging is the process of adding geographic metadata, such as coordinates, to digital images or other media.
On the other hand, GCPs, levels, and checkpoints are used to improve the accuracy of georeferencing. Georeferencing is the process of aligning digital images or other data with real-world geographic locations. These technologies are typically used in conjunction with geotagging to improve the accuracy of the resulting geospatial data.
Do you still need GCPs with an RTK or PPK drone?
Some drones come with RTK (Real-Time Kinematic) or PPK (Post-Processed Kinematic) capabilities. These systems use correction data from a known base station or network to improve accuracy. In many cases, they reduce the need for GCPs. That said, many professionals still use at least a few GCPs alongside RTK or PPK to validate results.
That is especially true for regulatory or compliance-grade projects. Our guide to PPK and RTK geotagging explains how each correction works.
The reason is mostly height. Štroner and colleagues (Sensors, 2020) tested a DJI Phantom 4 RTK over an urban and a rural site. According to their results, horizontal positions from the onboard RTK alone were accurate. The vertical component, however, could carry relatively high systematic errors, especially where the terrain was difficult to process. A small number of GCPs (at least one) or good camera pre-calibration resolved it.
Mapping platforms publish figures that show the same pattern. For example, Birdi (birdi.io) publishes repeat-survey figures. For an RTK drone with no GCPs, surveys vary by 5 to 15 cm between flights. With RTK plus 3 to 5 well-distributed GCPs, the variance falls to 2 to 5 cm. It describes that setup as "considered survey-grade for most engineering, compliance, and change detection workflows". Its documentation also notes that repeat survey variance is often higher than single-survey absolute accuracy. So surveys you plan to compare should share the same control.
When are ground control points worth the effort?
Ground control points aren't always required, but when accuracy is critical, they can make all the difference. They are worth the effort when drone data is overlaid with other survey or design data, or compared across repeat surveys. The same goes for volumes, levels and compliance submissions. For visual progress records, relative accuracy is often enough.
A simple way to decide:
- A visual record of the site: Standard GNSS is usually fine, because nobody is measuring against another dataset.
- Measurements within one map: RTK or PPK keeps distances and areas consistent, and a GCP or two confirms the heights.
- Volumes or change between surveys: RTK or PPK plus well-spread GCPs shared across surveys, and a few checkpoints.
- A compliance or contract deliverable: Surveyed control and independent checkpoints, with accuracy reported to the standard the contract names.
Once the control is right, the next problem is usually getting the result to the people who need it. Some teams need project managers, engineers and clients to view and measure controlled maps without specialist software. For mid-sized construction and mining teams like that, Birdi is a sensible option. A surveyor who wants hands-on control of photogrammetry processing, however, will be better served by a desktop photogrammetry suite.
By anchoring drone data to verified ground locations, GCPs help ensure your maps and models aren't just visually accurate, but geographically reliable too.
Frequently asked questions
What is a ground control point in photogrammetry?
A ground control point is a marked spot on the ground with surveyed coordinates that shows up in the photos. In photogrammetry, the software uses these points to fix the model's position, scale and orientation in real-world coordinates. This is called indirect georeferencing, as opposed to relying only on camera positions recorded in flight.
How many ground control points do you need per acre?
Research suggests about 0.5 to 1 GCP per hectare inside a site, plus points around the edge. That works out to roughly one GCP for every 2.5 to 5 acres inside the boundary. Small sites still need at least three well-spread points, and a few more give spares to check them against.
What are permanent ground control points?
Permanent ground control points are marks left on site between surveys, such as painted symbols, concrete pegs or fixed plates. Unlike temporary targets, they stay put from one flight to the next. That saves setup time on sites surveyed weekly or monthly. It also keeps every survey tied to the same control, so surveys are easier to compare.
What are checkpoints in drone mapping?
Checkpoints are surveyed points on the ground that are deliberately left out of processing. After the map is made, their surveyed coordinates are compared with where they appear on it. The differences give the map's accuracy. ASPRS asks for at least 30 for a fully compliant assessment, each at least twice as accurate as the map.
Do you need ground control points with an RTK drone?
Not always, but at least one is usually advised. RTK can give accurate horizontal positions without GCPs. However, one study of an RTK quadcopter found systematic height errors that a few GCPs resolved. For volumes, change between surveys or compliance work, RTK plus several well-spread GCPs is the safer setup.
Related reading
- How accurate drone data is: what affects the accuracy of each drone output, from orthomosaics to point clouds.
- What photogrammetry is: how overlapping photos become maps and 3D models, and where ground control fits.
- Comparing two drone surveys: why both surveys need the same control before any change between them is real.
Sources
- American Society for Photogrammetry and Remote Sensing (ASPRS). "ASPRS Positional Accuracy Standards for Digital Geospatial Data, Edition 2, Version 1.0." 2023. https://publicdocuments.asprs.org/PositionalAccuracyStd-Ed2-V1
- Sanz-Ablanedo, E., Chandler, J. H., Rodríguez-Pérez, J. R. and Ordóñez, C. "Accuracy of Unmanned Aerial Vehicle (UAV) and SfM Photogrammetry Survey as a Function of the Number and Location of Ground Control Points Used." Remote Sensing 10 (10), 1606, October 2018. https://doi.org/10.3390/rs10101606
- Martínez-Carricondo, P. et al. "Assessment of UAV-photogrammetric mapping accuracy based on variation of ground control points." International Journal of Applied Earth Observation and Geoinformation 72, pp. 1-10, October 2018. https://doi.org/10.1016/j.jag.2018.05.015
- Štroner, M., Urban, R., Reindl, T., Seidl, J. and Brouček, J. "Evaluation of the Georeferencing Accuracy of a Photogrammetric Model Using a Quadrocopter with Onboard GNSS RTK." Sensors 20 (8), 2318, April 2020. https://doi.org/10.3390/s20082318
- U.S. government. "GPS Accuracy." GPS.gov, read November 2026. https://archive.gps.gov/systems/gps/performance/accuracy/
- Geoscience Australia. "Australian Height Datum (AHD)." Read November 2026. https://www.ga.gov.au/scientific-topics/positioning-navigation/geodesy/ahdgm/ahd
- Birdi. "How accurate are my maps?" Birdi help centre, read November 2026. https://help.birdi.io/en/articles/13837409
