Ground Sample Distance (GSD) Calculator

Background.

Ground sample distance is the number that decides whether a survey flight is worth flying. It is the ground distance one image pixel covers — 2.5 cm per pixel, 30 cm per pixel, 30 metres per pixel — and it is the single figure clients specify, deliverables are judged against, and flight plans are built around. ASPRS defines it in the Positional Accuracy Standards for Digital Geospatial Data as "the linear dimension of a sample pixel's footprint on the ground". Pix4D puts the same idea operationally: the GSD is "the distance between two consecutive pixel centers measured on the ground", so a GSD of 5 cm means one pixel represents 5 cm linearly and 25 square centimetres of area.

The arithmetic is similar triangles and nothing more. A pixel of physical size p sitting at focal length f behind a lens sees a patch of ground p × H ÷ f across, where H is the height above that ground. Divide the sensor width by the image width to get the pixel pitch, multiply by the height, divide by the focal length, and you have the answer. This calculator will solve that identity in any of three directions, because in practice you rarely know the same two things twice: how sharp will imagery from this height be, how low must I fly to hit a target, and — the one most calculators skip — what lens do I need when regulation pins my altitude at 120 metres and the client still wants 2 centimetres per pixel?

Alongside the GSD it returns the frame footprint, because that is what turns a resolution figure into a flight plan. A useful property falls out of the algebra: the footprint is sensor width × height ÷ focal length, with the pixel count cancelling out entirely. Doubling your camera's megapixel count halves the GSD but does not widen the frame by a millimetre. That is why a higher-resolution sensor lets you fly higher for the same GSD and therefore cover more ground per battery, and it is the single biggest lever on survey productivity.

The page also applies the one hard rule the standards impose. ASPRS Edition 2 states that when producing digital orthoimagery, the GSD as acquired "should not be more than 95% of the final orthoimage pixel size" — so a 2.74 cm raw GSD supports an orthomosaic no finer than 2.89 cm per pixel. Delivering a 2.5 cm orthomosaic from that flight would be resampling, not resolution, and this calculator says so rather than letting you find out at acceptance.

Two cautions run through everything below. GSD is computed at one reference plane, and ASPRS's own definition warns that in raw imagery "pixel size is not uniform and varies based on sensor orientation and terrain" — fly 100 m above a valley floor and a ridge 30 m up is sampled 30 % finer. And GSD is a sampling interval, not a measure of sharpness and certainly not a measure of accuracy: a motion-blurred, poorly controlled dataset has exactly the GSD its geometry says it has. Below the widget you will find the identity derived, a worked example computed by hand against a published drone specification, and an account of where the model stops being true.

What is ground sample distance calculator?

Ground sample distance is the spacing, projected onto the ground, between the centres of two adjacent pixels in an image. It is the standard way of describing the spatial resolution of aerial and satellite imagery, and it has a formal definition: ASPRS's Positional Accuracy Standards for Digital Geospatial Data, Edition 2, defines it as "the linear dimension of a sample pixel's footprint on the ground", adding that "in raw imagery, pixel size is not uniform and varies based on sensor orientation and terrain" and that "the term 'nominal GSD' refers to the average or approximate size of pixels in raw imagery", while "in orthorectified imagery, the GSD for all pixels is uniform and constant regardless of the terrain variation". That distinction matters commercially: the GSD written into a contract is almost always the orthoimage pixel size, which is a processing output, while the GSD a flight plan controls is the nominal raw GSD at mean terrain. The scale spans an enormous range. A drone at 2 metres inspecting a weld operates at fractions of a millimetre per pixel; a mapping flight at 100 metres is at a few centimetres; the USGS Landsat 8 OLI instrument delivers 15-metre panchromatic and 30-metre multispectral imagery, with its thermal bands at 100 metres. Every one of those figures comes from the same relation between pixel pitch, focal length and distance. What GSD is not is resolution in the optical sense. Resolution is about whether two nearby features can be told apart, which depends on the lens, on focus, on diffraction, on motion during the exposure and on atmospheric conditions. GSD only says how finely the scene is sampled. A perfectly focused image and a badly blurred one taken from the same height with the same camera have identical GSD. Nor is GSD accuracy: ASPRS Edition 2 moved deliberately to accuracy thresholds "independent of published GSD", precisely because a fine GSD with poor ground control produces detailed imagery in the wrong place.

How to use this calculator.

1. 01Choose what you are solving for. "GSD" answers how sharp imagery from a planned height will be. "Altitude" answers how low to fly for a client-specified GSD. "Focal length" answers which lens meets the spec when the altitude is fixed by regulation or by terrain.
2. 02Enter the sensor's imaging-area width in millimetres. If the manufacturer publishes only an actual and a 35 mm-equivalent focal length, divide the equivalent by the actual to get the crop factor, then divide 43.27 mm by that to get the sensor diagonal — the hint on the field walks through this for the default camera.
3. 03Enter the still-image width and height in pixels. Use the stills resolution, not the video resolution: many cameras crop the sensor for video, which would make the pixel pitch wrong.
4. 04Enter the lens's actual focal length in millimetres, never the 35 mm equivalent. The sensor width already accounts for the format, so an equivalent focal length would apply the crop factor a second time.
5. 05Enter the height above the ground being imaged. This is height above terrain. Most autopilots hold height relative to the take-off point, so on sloping ground the GSD you actually acquire drifts across the site unless terrain-following is switched on.
6. 06Read the GSD, then read the frame footprint. Divide the footprint dimensions by your overlap requirements to get the shutter interval along track and the flight-line spacing across track — that is the step that turns a resolution number into a mission.
7. 07Check the smallest orthoimage pixel figure against what the deliverable specifies. If the contract asks for a finer orthomosaic than this flight supports under the ASPRS 95 % rule, you need to fly lower, use a longer lens or use a higher-resolution sensor — resampling will not create detail that was never acquired.
8. 08Treat the result as a nominal figure at one plane. Add margin for terrain relief, and remember that GSD says nothing about whether the imagery is sharp: check your shutter speed against the aircraft's ground speed separately.

The formula.

The derivation is a pinhole and a pair of similar triangles. Put the lens at the apex. On the image side, a feature of size p sits at distance F from the apex; on the object side, its counterpart of size G sits at distance H. The two triangles share an angle, so G/H = p/F, which rearranges to

G = p × H / F

Take p to be the width of one photosite and G is the ground sample distance. The pixel pitch itself comes from the two figures every camera publishes:

p = sensor width (mm) / image width (px)

For the Phantom 4 RTK, 13.2 mm ÷ 5472 px = 0.0024122807 mm, or 2.4123 µm per pixel. At 100 m with the 8.8 mm lens:

GSD = 0.0024122807 × 100 / 8.8 = 0.0274122807 m = 2.7412 cm per pixel

The two rearrangements the other modes use follow immediately: H = GSD × F ÷ p, and F = p × H ÷ GSD.

The frame footprint is the same relation applied to the whole sensor rather than one pixel:

footprint width = SW × H / F footprint height = SH × H / F

Notice what dropped out. Multiply GSD by the image width and the pixel count cancels: GSD × ImW = (SW/ImW) × H/F × ImW = SW × H / F. The footprint depends only on the sensor's physical size, the height and the focal length. Doubling the megapixel count of the same sensor halves the GSD and leaves the footprint untouched. This is the arithmetic behind the productivity argument for higher-resolution survey cameras: a finer GSD at the same height, or the same GSD from a greater height, which means fewer flight lines and fewer batteries for the same site.

One externally-imposed number appears on the page, and it is quoted rather than invented. ASPRS Positional Accuracy Standards, Edition 2, section B.2: "When producing digital orthoimagery, the GSD as acquired by the sensor (and as computed at mean average terrain) should not be more than 95% of the final orthoimage pixel size." Rearranged, the smallest orthoimage pixel a given raw GSD can support is GSD ÷ 0.95. For the worked example, 2.7412 ÷ 0.95 = 2.8855 cm per pixel.

What the model deliberately leaves out: terrain relief, which makes the acquired GSD vary continuously across a site and is the reason ASPRS distinguishes nominal from actual; camera tilt, since an oblique frame has a GSD that changes from the near edge to the far edge; lens distortion, which photogrammetric processing corrects but which does make raw pixel geometry non-uniform; and image sharpness, which is a completely separate question governed by focus, diffraction, and the shutter speed against your ground speed.

A worked example.

Frequently asked questions.

References& sources.

1. [1]American Society for Photogrammetry and Remote Sensing (2023). ASPRS Positional Accuracy Standards for Digital Geospatial Data, Edition 2, Version 1.0. Definitions: 'ground sample distance (GSD) – The linear dimension of a sample pixel's footprint on the ground. In raw imagery, pixel size is not uniform and varies based on sensor orientation and terrain. The term "nominal GSD" refers to the average or approximate size of pixels in raw imagery. In orthorectified imagery, the GSD for all pixels is uniform and constant regardless of the terrain variation.' Annex B.2: 'When producing digital orthoimagery, the GSD as acquired by the sensor (and as computed at mean average terrain) should not be more than 95% of the final orthoimage pixel size.' Linked here via a hosted copy because asprs.org returns HTTP 403 to automated requests; the quoted text was read directly from this PDF.
2. [2]Pix4D SA. Ground sampling distance (GSD) in photogrammetry — product documentation. 'The Ground Sampling Distance (GSD) is the distance between two consecutive pixel centers measured on the ground', and 'A GSD of 5 cm means that one pixel in the image represents linearly 5 cm on the ground (5*5 = 25 square centimeters)'. Uses the same variable set as this calculator: flight height H, image width ImW, sensor width SW, focal length F.
3. [3]DJI. Phantom 4 RTK — Specs. Source of every default on this page: camera sensor '1" CMOS; Effective pixels: 20 M'; lens 'FOV 84°; 8.8 mm / 24 mm (35 mm format equivalent: 24 mm); f/2.8 - f/11'; still photography image size '4864×3648 (4:3); 5472×3648 (3:2)'. The 13.2 mm sensor width used as a default is derived from the published 8.8 mm / 24 mm-equivalent pair, not quoted from DJI.
4. [4]U.S. Geological Survey. Landsat 8 mission page. The Operational Land Imager collects data at '15-meter panchromatic and 30-meter multi-spectral spatial resolutions', and the Thermal Infrared Sensor's two bands are collected at 100 m. Used on this page as the scale anchor for comparing drone-scale and satellite-scale ground sample distances.