PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
December 2020
721
Question One: I was asked to provide services for surveyors, which
of course means that absolute accuracy is key. The surveyors flew
a sample project and provided me with a list of GCP coordinates.
They also hired a company that specializes in processing images for
surveyors to verify and compare data. The other company was able
to process the data using a leading UAS data processing software
and get the residuals down to under 0.01-0.02 ft. However, when
I processed the same data using the same GCPs and the same
software, I could not get the RMSE much lower than 0.07 ft. Digging
into the problem further, I found that I needed to turn on the “rolling
shutter” option in the software since the camera has a rolling
shutter. Once I fixed that, my results matched the other company’s
results. Can you share your thoughts on that function? Also, in
your article, “Harnessing Drones the Photogrammetric Way,” you
mentioned that the RMSE figures from aerial triangulation should
not be used alone as proof of product accuracy. Can you elaborate?
Nathan Mangsen, Mangsen Mapping
Dr. Abdullah:
I am glad you mentioned the issue of the rolling
shutter, as it is important to the UAS mapping community.
Shutter plays a vital role in camera operation as it is used to
control the passage of light into the focal plane of a camera
resulting in a perfectly exposed image. Digital cameras are
manufactured with either mechanical shutters, electronic
shutters or a combination of the two. Consumer-grade
cameras, like those used with UAS, are designed according
to two mechanisms, the leaf shutter (also known as the
“dilatating aperture shutter”) or the rolling shutter (also
called “curtain” or “sliding shutter”). The leaf shutter, which
is usually mechanical, is favored in the mapping community
because it freezes the scene as it exposes the entire sensor
array. On the other hand, the rolling shutter exposes one
line of pixels at a time, creating a sliding curtain effect, until
the entire image is exposed. When a rolling shutter is used
on a camera mounted on a mobile platform like UAS, the
imagery is distorted and can blur. This distortion is exag-
gerated by high aircraft speeds or low light conditions. Most
image processing software that is designed for precision
mapping provides an option to correct for rolling shutter
distortion during data processing. For aerial mapping
from a mobile platform, such as a UAS, leaf shutter is
recommended because it minimizes or eliminates image blur
due to aircraft motion. Unfortunately, many UAS users are
not aware of this fact. For those who are aware, many do not
know that some (not all, unfortunately) UAS data process-
ing software provides the option to correct for the anomaly
caused by rolling shutters. This option needs to be activated
by the user during processing since the software does not
automatically know whether the imagery was collected
with a lens that has a rolling shutter. If the effect of the
rolling shutter is not addressed during data processing,
the positional accuracy of the product will be compromised.
Your question brings up a good example of this problem. My
suggestion for users of UAS data processing software is to
first figure out whether the camera that will collect the data
is equipped with a rolling shutter and if so, to figure out
whether the software provides an option to correct for the
rolling shutter effect.
Let me now address your question on basing final product
accuracy on the quality of the results of aerial triangulation
or, in more specific terms, the fit to the ground control points
(GCPs) during the aerial triangulation process. There are
several reasons why we should not use the estimated errors
of aerial triangulation as final indicators for derived product
accuracy. Among them are:
1.
Survey checkpoints should be independent from
the production process:
The correct way to evaluate
the positional accuracy of any geospatial product is to
use an independent set of survey checkpoints that were
“If the effect of the rolling shutter is not addressed
during data processing, the positional accuracy of
the product will be compromised.”
Photogrammetric Engineering & Remote Sensing
Vol. 86, No. 12, December 2020, pp. 721–722.
0099-1112/20/721–722
© 2020 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.86.12.721
