There are a variety of factors and considerations to keep in mind that affect the accuracy of your HPP data collection. This article will cover definitions of basic terms when discussing accuracy, how they relate to one another, and performance and other impacts of the Skycatch RTK system.
Covered in this article:
- Definition of Accuracy, Resolution/Spatial Resolution, and GSD
- The relationship between these three items
- How GSD of inputs affects accuracy of 3D outputs
- Choosing flight height
- Performance of the Skycatch RTK system
- Other impacts on accuracy of Skycatch RTK system
- Accuracy: the amount which an identifiable object or position in output data matches the real-world position of that object. For example, when we say we can achieve 3 cm Z accuracy, it means the elevation of a surface as measured in our output data (point cloud, surface) is within 3 cm of that surface if a surveyor was to go measure it using their established best practices (GPS or total station, for example).
- Resolution: this can be thought of as the density of information in a piece of data. When the term resolution is used in drone data, it is usually referring to the “spatial resolution." For example, we might say that the resolution of an orthophoto is 5 cm. More properly put, the “spatial resolution” of the orthophoto is 5 cm per pixel. So an object that is 10 pixels wide in the orthophoto is 50 cm long.
- Spatial resolution may apply to 3D data as well such as point clouds. For example, a Skycatch point cloud may contain up to 100 points per square meter.
- GSD: this is the spatial resolution of the photos that the drone takes to make a map. A smaller GSD (e.g.- 1 cm) means a more detailed image, and a larger GSD (e.g.- 5 cm) means a less detailed image. The GSD is dependent upon the drone’s flight height above ground: GSD increases as height increases.
GSD increases as the drone height increases because the ground is further away from the drone’s camera. When GSD increases, spatial resolution usually increases too. And when resolution increases, i.e. gets worse, accuracy decreases. So as a general rule, you can achieve higher accuracy when flying lower, and lower accuracy when flying higher.
As a general rule, the 3D accuracy of an output is about 2-3 times the GSD of the input data. This is due to the way photogrammetry uses triangulation to calculate elevations. This is true for other types of systems: for example, survey GPS Z-accuracy is about 2-3 times the X-Y accuracy.
So if you have a GSD of 2 cm, you can expect the output Z-accuracy to be from 4-6 cm — as long as all other inputs are normal (images are sharp, good lighting, RTK fix, etc. Some of these are covered later).
The trade off with flying at a lower or higher altitude is data collection rate. It might take you 2 hours to fly a site at 40m that you could fly at 120m in a half hour.
For example, a 50 acre plot can be mapped by the Skycatch drone in one 20-minute flight at 120m, the legal maximum in many countries. This will achieve a GSD of approximately 4 cm, which results in a Z-accuracy of 8-12 cm. For many purposes, this is useful 3D data. But if you need to produce sub-5 cm accuracy (0.1 ft), you could fly at 50m. The High Precision Package GSD at 50m is about 1.6 cm, so we would expect the accuracy to be between 3.2 cm and 4.8 cm.
If you want to achieve higher accuracy, you can fly lower — to a point. The limits have to do with the RTK system.
With a good RTK fix, the Skycatch system achieves photo positioning accuracy of about 2 cm in X-Y and 3 cm in Z. This is similar to survey GPS and is quite a feat given that it’s triangulating from satellites 20,000 km in space! When a dataset's 3D accuracy based on GSD (i.e. 2 cm GSD @ 60 m x 2-3 = 4-6 cm) is larger than the 3 cm Z accuracy typical for RTK, photogrammetric accuracy is the limiting accuracy. When the dataset’s 3D accuracy based on GSD (i.e. 0.5 cm GSD @ 20 m x 2-3 = 1-2 cm) is smaller than the Z accuracy typical for RTK, RTK position accuracy is the limiting accuracy.
For a combination of the reasons above, a 60 m altitude is a good “sweet spot” of low GSD, high accuracy, and good area coverage. Extensive testing has shown that flying at 60 m yields an average Z accuracy of 5 cm, right as expected: 60 m GSD = 2 cm x 2-3 = 4-6 cm with an average of 5 cm.
In our testing, Skycatch has seen that the accuracy of using a virtual reference station, or RTK network, is approximately the same as using a local base station to take static observations for correction. Specifically, image positions determined with VRS/nRTK are within 1 cm of positions determined with a local base. So, it’s generally ok to simply choose the correction method that is most convenient for you on site.
As a rule of thumb, RTK fix accuracy is affected as 1ppm of distance to the base. For example, when using a local base station, 1 mm of variance is added to the solution for every km the drone is flying from the base. So, large sites can be mapped using one base: flying anywhere within a 5 km radius of the base will result in only 5 mm of additional potential error in the solution.
Generally, this means that the need to relocate the base or VRS position would be required only when flying long, linear projects over 10 km.