Drone photogrammetry relies on using imagery taken by a drone to create insights for users which wouldn’t otherwise be available. In this article we’re specifically dealing with RGB photogrammetry as opposed to LIDAR or multispectral.
Chasing insights
Sometimes the task is to simply create an orthomosaic1 showing a RGB image of the terrain. This is equivalent to “I want an aerial picture of what’s there”. In many circumstances that is all that’s needed, a bird’s eye view of the terrain.
Often however the intent is to generate some insight into the site. This is where an RGB orthomosaic may not be the best way to present the data.
Creating the right picture
Let’s consider the picture above. It consists of 4 layers. Starting from the “bottom” - that is the furthest back layer these are:
- Base map - this helps to situate the picture in the landscape;
- Digital elevation model (DEM);
- Another DEM;
- Contours - in this case at 25cm intervals
About Surfaces
Firstly let’s talk about the DEM. In this case, because the underlying data is a RGB orthophoto, the DEM is a surface model, more than it’s a terrain model. To explain, if the data was gathered using a LIDAR then there’s a good chance that the “surface” would be the ground or near to it, therefore a terrain model. This is because of LIDAR’s capacity to penetrate vegetation, particularly the tree canopy2. In our case the RGB imagery can’t penetrate the vegetation with any real effectiveness. This means that, even though we seek to classify the data to ground, there’s effectively no accurate data about ground level within areas covered with trees and scrub. Hence we are dealing with a digital surface model (DSM). In this image the important area is the steeply sloped area which is free of vegetation and therefore RGB data is sufficient.
Why multiple DEMs?
Layers 2 and 3 in the picture are both DEMs - in this case DSMs. Why do we need two of them? The answer is all about achieving our goal of providing easily interpreted insight. Photogrammetry is only useful if the viewer can easily discern the detail they want.
A DEM can be coloured in any way the software is capable of. In our case the DEM comes in raw as a picture shaded from black to white. It’s hard to interpret and not useful. Instead we’ve taken layer 2 and used “hillshade” to elicit detail from the picture.
The hillshade brings the DEM to life and gives it shape and volume. Immediately it’s possible to see detail - lots of detail - which you can’t easily see in the raw DEM. However it doesn’t give us some information about height for instance or give us a clear understanding of slope. For that we can introduce a false colour DEM - layer 3:
Now the DEM is giving us some information about relative heights, but it’s lost of lot of the detail which we gained from the hillshade. So we combine the two layers:
The picture is now springing to life. The hillshade, combined with the colouring starts to create an easily interpretable picture. What’s still missing is detailed elevation data. For that we can use contours:
Again on their own, they’re interesting and useful. But when combined, with the other layers (as they are in the first picture) they really come to life.
Multiple outputs from the one flight
The great thing about RGB photogrammetry is that you can generate multiple outputs from the one flight. If you just want to know what the site looks like, then you have the simple RGB photomosaic.
RGB Photogrammetry - flexible and powerful
There’s a lot to be said for simple, effective RGB photogrammetry, particularly when you work with someone to really deliver the outcomes you need.
Footnotes
Explanations of many of the terms in this article can be found in our Glossary↩︎
If LIDAR is so effective at imaging the real surface, why wouldn’t we always use it? The answer is twofold: Often, particularly in areas without much overlying vegetation, RGB imagery gives just as good an outcome as LIDAR. LIDAR is also much more capital intensive to own and therefore the costs to collect LIDAR data are much more than those to collect RGB data. It comes down to delivering the best outcome for the best price.↩︎