Orthorectification: Turning Raw Satellite Data Into Map-Ready Imagery

Summary

Orthorectification is the geometric correction that moves every pixel in a satellite image to its true ground position. It removes displacement caused by terrain relief, sensor view angle and the curvature of the Earth, using the image’s rational polynomial coefficients together with a digital elevation model. The corrected output is an orthophoto: uniform scale throughout, and measurable like a map.

• Raw imagery is not a map. Objects sit away from their real positions, by metres on flat ground and by considerably more in steep terrain.
• The correction needs two inputs beyond the image: RPCs, which are supplied in the image metadata, and a DEM, which you have to source.
• You can run it yourself with GDAL’s gdalwarp. The command is one line. Choosing the right DEM and validating the result is the actual work.
• Every image Geopera delivers arrives already orthorectified, so the question usually does not come up.

When satellites orbit hundreds of kilometres above Earth, they capture a lot of detail. What comes straight off the sensor is not a map. Raw satellite images carry distortions from terrain, sensor tilt, and the curvature of the planet itself.

Orthorectification removes those distortions. It aligns each pixel to its true ground position and turns the image into something you can measure against.

Why Orthorectification Matters

A satellite is not always looking straight down. The sensor has a view angle, the ground has hills and valleys, and the planet curves away underneath. Each of those shifts objects away from where they actually are.

Without correction:

• Features do not appear in their real locations.
• Distance and area measurements are unreliable.
• Overlaying other geospatial data produces visible misalignment.

The size of the error scales with relief and with how far off nadir the sensor was looking. On flat ground at near-nadir it can be small enough to ignore. On a steep slope at a high view angle it is large enough to put a feature on the wrong side of a boundary.

For mining, infrastructure, agriculture and environmental monitoring, that matters. Engineers cannot measure, planners cannot design against it with confidence, and analysts risk drawing conclusions from imagery that disagrees with the ground.

Orthorectification produces an orthophoto, where every pixel sits in its proper place on Earth.

How Orthorectification Works

Three inputs go in:

• Raw image data, the original satellite capture.
• RPCs (Rational Polynomial Coefficients), metadata that describes the relationship between image pixels and ground coordinates.
• A DEM (Digital Elevation Model), which supplies terrain height.

They come together in three steps. The RPCs give a mathematical model linking each pixel to a location on Earth. The DEM then adjusts those positions for elevation, so that slopes, valleys and ridges land where they belong rather than where a flat-earth assumption would put them. Finally the imagery is resampled onto a consistent grid in a chosen map projection.

The result is an image whose spatial relationships hold up: features align with maps, site plans and other geospatial layers.

Doing It Yourself with GDAL

The open-source GDAL library will do this. gdalwarp applies the image’s RPCs and a DEM:

gdalwarp -rpc -to RPC_DEM=dem.tif raw_image.tif ortho_image.tif

-rpc applies the Rational Polynomial Coefficients from the image metadata. -to RPC_DEM=dem.tif names the terrain model used for elevation correction. The output file is the orthorectified product.

That command is the easy part. The work is in everything around it: picking a DEM whose resolution and vintage suit the terrain, choosing a resampling method that does not smear detail you need, validating the result against known ground positions, and running it over datasets large enough that memory and I/O start to matter. A coarse or outdated DEM will happily produce a confidently wrong orthophoto.

If you need elevation data to try this, the free sources of DEM data guide covers where to get it.

Ordering Orthorectified Imagery

If you would rather not run the pipeline, orthorectification is something you can buy already done.

Every image Geopera delivers is orthorectified before it reaches you, using our own elevation models or your supplied DEM and ground control where you have them. It arrives projection-ready for GIS, CAD and analytics platforms.

Orthorectification is one stage in a longer chain. Ingest, co-registration, ground control and ortho, pansharpening, atmospheric correction, cloud and shadow masking, seamline optimisation, harmonisation and QC all run as one automated flow, with nothing manual between stages and nothing hand-finished. That is the Legato process, and it is why imagery from different sensors and different dates arrives in a state where you can compare it directly. Delivery is under 24 hours.

You can search the archive and see prices per area at portal.geopera.com. For what the resolution tiers cost, see the satellite imagery cost guide.

Common Questions

What is the difference between georeferencing and orthorectification?

Georeferencing assigns map coordinates to an image. Orthorectification goes further and removes the geometric distortion first, using a terrain model, so that scale is uniform across the whole frame. A georeferenced image can still be wrong in the hills. An orthorectified one has had that displacement taken out.

What is an orthophoto?

An orthophoto is the output of orthorectification: an aerial or satellite image corrected so that it has the geometric properties of a map. You can measure distance and area off it directly.

Do I need orthorectified imagery?

If you are measuring anything, overlaying imagery on other spatial data, or comparing captures from different dates or sensors, yes. If you only need a visual impression of a location, raw imagery may be adequate. Anything that ends up in front of a regulator, an engineer or a court should be orthorectified.

Does orthorectification fix everything?

No. It corrects geometry, not radiometry. Haze, illumination differences between dates and colour mismatches between sensors survive it untouched, which is what atmospheric correction and harmonisation deal with later in the chain. It also has a known limit in dense urban areas: a standard terrain model corrects the ground, so the bases of tall buildings land correctly while their tops still lean.

The Takeaway

Raw satellite data is not reliable for precision work until it has been orthorectified. The process makes images line up with the real world, which is what makes measurement and change detection defensible.

You can do it yourself with GDAL. If you would rather receive imagery that is already corrected, aligned and map-ready, Geopera delivers it that way by default.