Share on facebook
Share on twitter
Share on linkedin

How ground sample distance (GSD) relates to accuracy and drone ROI

Ground sample distance (GSD) is a key limiting factor in the accuracy of your drone mapping project. At the same time, GSD influences how many GB of data you gather per survey and how long a project takes. 

GSD: the ground according to one pixel

The size of the real world captured by a pixel in a digital image, GSD is based on the flight altitude and camera parameters, which we discuss below. When you understand GSD as a cm per pixel relationship, it’s easy to calculate the size of the ground and features captured in a drone image.

For example, if your GSD is 1 cm (0.4 in), and you zoom into an object measuring 50 pixels across, you can be sure that object measures approximately 50 cm (19.7 in).

GSD is the amount of actual ground captured by the distance between the center point of two adjacent pixels.

Accuracy and drone ROI

Let’s expand on this logic and say you take aerial images of a mine site featuring some pipelines. You set your flight and payload parameters to achieve a GSD of 2.5 cm (1 in). On the resulting drone map, you want to find the length of one stretch of pipeline. To do this, you draw a polyline in your software along the distance of the pipe. The software will convert the line’s pixel distance into a real-world distance.

On the other hand, if you want to know the length down to the centimeter (0.4 in) level, you cannot measure it exactly. Why? Any space smaller than a 2.5 cm (1 in) will be contained within one pixel, which is a single square of data that lacks definition. The end points of your polyline will be sitting at best within 2.5 cm from the end of the pipe.

These aerial images by WingtraOne with its Sony QX1 payload compose a map of a Swiss mine based on a 2.5 cm (1 in) / px GSD. Zooming in to the closest point, we can get accuracy within a few centimeters (about an inch) to estimate the length of the pipe.

The same is true if you need to accurately locate the center of a ground control point or the edge of a building. With a better (lower) GSD, it will be much easier to see the center point or edge, because each pixel represents a smaller space. In the context of your survey, this kind of detail difference adds up over large areas, and it matters.

A 1 cm (0.4 in) error spread over many meters becomes a volumetric or land area difference that could get quite expensive.

For example, if our GSD is 5 cm (2 in), this means our accuracy error will be that much, or more, in cm. For a volume it’s actually way more than this. Why? You will factor this 5 cm (2 in) error into the surface area of X cm/in2 based on points plotted and the actual edge of the stockpile. So there will be at least this much difference between the surface area estimate and the actual edge of the stockpile, and this square surface area difference will be multiplied (cubed), as a volume. Imagine a 5 cm (2 in) thick blanket (of error) lying on top of the volume itself. The volume of that blanket depends on your GSD and accuracy.

How surface accuracy impacts volume accuracy

Actual stockpile

Lower accuracy

High accuracy

Stockpile measurement depends on a tight reading around a pile’s surface. The better (lower) your GSD is, the better your accuracy when measuring this surface, which reduces the difference between the real volume and the measured volume to a minimum.

All of this is straightforward when you are aiming toward relative accuracy, i.e., when the distance between any two points on a map reflect a real-world distance at the same scale. However, when you need absolute accuracy—where every point on the map is related to scale and is geotagged precisely to the real point on Earth—your GSD means something more.

What GSD and accuracy levels do you actually need?

Accuracy ranges from relative to survey-grade-absolute.* Absolute (ground-based) GPS data is mapped onto every pixel of every image of an aerial map. So every pixel represents a real GPS location.

Relative accuracy

If a map represents three meters in a centimeter, every centimeter on that map ideally measures three meters in real life. The amount that the map distances differ from this exact real-life scale indicates its relative accuracy.

Absolute accuracy

Absolute accuracy automatically accounts for relative accuracy and adds a level of precision. Each point on the map relates to an exact (X longitude, Y latitude, Z altitude) geo-point on/above the Earth’s surface.

So if you have a GSD of 1 cm (0.4 in), a high-end drone with a high-quality payload and multi-frequency GNSS receiver to correct for the precise position can give you an absolute accuracy of 1 cm (0.4 in). In contrast, an average drone gives you an absolute accuracy of up to 3 cm (1.2 in). These differences become more exaggerated as GSD gets worse (higher). A GSD of 10 cm (3.9 in), for example, will give you an absolute position accuracy of between 10 and 30 cm (3.9 and 11.8 in), depending on the quality of your drone and payload.

As a rule of thumb, you will be able to get measurements of absolute location at an accuracy of one to three times your GSD.

To get a lower (better) GSD, and a more accurate map, you need to fly lower (relative to your own camera resolution) and take more and higher-resolution pictures. This means more time spent on a project due to longer flights and more data to process. For example, if you fly with a GSD of 1 cm (0.4 in) instead of 3 cm (1.2 in), you are capturing and storing up to ten times more data for the same area covered. If your project does not depend on super-high accuracy, choose a higher GSD for improved speed, keeping in mind the absolute accuracy can be up to 3x the GSD.
Lower GSD Higher GSD
High accuracy
Low accuracy
Lower flight altitude
Higher flight altitude
Longer flight time
Shorter flight time
Up to 10x more data to process
Still a lot of data, but not extreme

How GSD relates to flight and project parameters. This table assumes the same area covered and the same sensor. If you have a higher resolution sensor, your flight time will decrease for the same GSD and area.

Why choose a high-quality drone?

Remember that GSD is directly related to the number of pixels in a single shot. So with a high-resolution, 42MP sensor, you will fly 1.5x higher and cover 1.5x more ground in the same time as a 20MP sensor flying with the same GSD setting. High-end, heavier sensors require a high-quality VTOL drone to carry them over a broad area, safely.

Add to this, when you need details down to a centimeter or a half an inch, your results also depend on the size of the drone camera’s sensor. In lower-quality drones, you’ll find their payloads have cropped sensors (see sensor size diagram). Larger sensors capture more light in less time—they enable a faster shutter speed to achieve sharp, well-exposed results, which improves the accuracy.

Camera sensor size comparison between Full-frame WingtraOne + RX1Rll, Aeria X, 1 inch senseFly S.O.D.A., DJI Phantom
The size of a payload’s sensor impacts the quality of images, especially those taken from a lower altitude and at high speeds. Measuring the real resolution of images with the same GSD gives an idea of the difference a high-quality payload makes.

To get a better idea of this, visit our drone comparison reports to see the real resolution differences across images taken by cropped and full-frame payloads.

Conclusion: What we recommend

In the end, for most professional surveyors, we recommend a GSD of 1 cm (0.4 in) or less in combination with PPK data. This way, project accuracy stays below 3 cm (1.2 in) absolute accuracy.

If you want to maximize the accuracy of your projects at this GSD, choose a drone with a high-quality, multi-frequency GNSS receiver. This supports absolute accuracy closer to your GSD based on a payload with a higher resolution. An example of this would be the WingtraOne with its high-quality PPK option and RX1R II 42MP full-frame payload.

*Survey-grade accuracy must be so precise that it could be used in a court of law as back up data supporting claims for things like land ownership or placement of boundaries. For example, standards around survey-grade accuracy are detailed here for the US. Flood-prevention and cadastral studies are two examples of where survey-grade accuracy is mandated, and this data must often be stamped by a surveyor licensed to provide such data.

Share on facebook
Share on twitter
Share on linkedin

Published by

Wingtra
Wingtra

Wingtra develops, produces and commercializes high precision VTOL drones that collect survey-grade aerial data.

Read about our drone

Read more

We use cookies to provide a user-friendly experience. By continuing to browse this site, you give consent for cookies to be used and stored on your device. To find out more please read our Privacy Policy.