True Orthophoto vs Orthophoto: Why Buildings Lean and How to Stop It
An orthophoto is corrected so that it can be measured like a map, but the correction is only as good as the surface it uses. Rectify a photo on the bare ground and buildings, trees and stockpile edges still lean. A true orthophoto is rectified on a surface that includes them, so roofs sit on their footprints and the ground behind them is filled from photos that actually saw it.
Relief displacement: why tall things lean
A camera sees in perspective. A point above the ground appears further from the centre of the photo than the ground directly below it, and the shift grows with the object's height and its distance from the photo centre and shrinks as you fly higher. In the classic form, the displacement on the image is the radial distance times the object height divided by the flying height.
Orthorectification removes that shift by projecting each photo onto a surface model. Whatever the surface knows about is corrected. Whatever it does not know about keeps its lean: on a terrain model with no buildings in it, a roof is drawn displaced from its walls and covers ground it should not.
Orthophoto and true orthophoto
A conventional orthophoto is rectified on a terrain model of the bare ground. Ground features are in the right place, but anything that stands above the ground leans away from each photo's centre, and the lean changes direction from one photo to the next, which is what makes building edges look torn along mosaic seams.
A true orthophoto is rectified on a surface model that carries the height of the ground and of what stands on it. Roofs are drawn over their footprints, walls disappear, and ground that one photo could not see behind a building is taken from another photo that could. That last step is why a true orthophoto needs plenty of overlapping views of every cell.
| Orthophoto | True orthophoto | |
|---|---|---|
| Rectified on | A terrain model of the bare ground | A surface model with buildings, trees and stockpiles |
| Tall objects | Lean away from the photo centre | Drawn upright over their footprint |
| Ground hidden behind objects | Covered or duplicated | Filled from a photo that saw it |
| Data needed | A terrain model | A surface model and high overlap |
| Where the difference shows | Little on open, flat ground | Buildings, structures, tall vegetation, stockpile edges |
Why drone photogrammetry makes true orthophotos practical
Producing a true orthophoto used to mean getting a detailed surface model of every building from somewhere else. A drone flight brings its own: the dense point cloud from the same photos gives a surface model of the ground and everything on it, and the high overlap of a mapping flight means most cells are seen by several photos. The surface and the mosaic come from the same data, so they agree.
Moving objects and ghosting
A mosaic has a second, separate problem: things that move between photos. A car, a person or a boat that sits in different places in two overlapping photos turns into a half-transparent double image if the photos are simply averaged. A true orthophoto does not fix this by itself, because it is about geometry, not about time.
The fix is in the blend. Where the overlapping photos disagree in a cell, take that cell from one photo instead of averaging it, and keep the even tone by blending only the low-frequency colour. Seamlines that avoid crossing buildings and moving objects help as well.
How Stream Capture makes the orthophoto
Stream Capture projects the photos onto the surface built from its own dense point cloud, so relief and what stands on the ground are corrected rather than smeared. The default blend measures how much the overlapping photos disagree in each cell; cells that disagree are taken from the photo that sees them best, while low-frequency colour is blended across photos for an even tone. Minimum-cost seamlines are available as an option, and a source-map raster records which photo each area came from.
There is also a quick-map mode that builds the surface from the sparse alignment points to answer whether the data is sound before you leave the site. It is labelled as not true-ortho, because on a coarse surface building and tree edges can lean.
Frequently asked questions
What is a true orthophoto?
An orthophoto rectified on a surface model that includes buildings, trees and other objects, not only the bare ground, so tall objects are drawn upright over their footprints and hidden ground is filled from other photos.
Why do buildings lean in an orthophoto?
Because the photo was rectified on a terrain model that does not contain the buildings. Their height is not corrected, so they keep the perspective lean of the original photo, away from its centre.
Is an orthomosaic a true orthophoto?
Not necessarily. Orthomosaic only says the image is assembled from many photos. Whether it is a true orthophoto depends on the surface it was rectified on.
Do I need a true orthophoto?
On open, flat ground the difference is small. Where there are buildings, structures, tall vegetation or steep stockpile edges, and especially if you will digitise edges from the image, a true orthophoto is the one that lines up with reality.
Can a drone produce a true orthophoto?
Yes. The dense point cloud from the same flight gives the surface model, and the high overlap of a mapping flight gives several views of most cells.
Why do moving cars appear twice on my orthomosaic?
The car was in different places in overlapping photos and the photos were averaged. A blend that takes disagreeing cells from a single photo avoids the double image.
Make a true orthophoto from your own drone photos
Native Windows · fully offline · free beta