PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
June 2019
407
Question:
I am a college student working on my bachelor degree in Spatial Science
(Surveying).
I am interested in photogrammetry and my study is on drone-based surveys. I have
the following questions on the new “ASPRS Positional Accuracy Standards for
Digital Geospatial Data:”
1. In sections 7.7 and 7.8, when it talks about checkpoint and ground control accu-
racy being 1/2 RMSEmap, is RMSEmap the desired/intended accuracy class?
2. With GCPs having three times the accuracy of the geospatial data set being
tested, does that mean the GCP accuracy will be three times more accurate
than the desired/intended accuracy class?
3. Do you use Table D.1 to calculate all the statistics and then use the results to
determine the ASPRS accuracy class? Is that the typical workflow? Is there a
sample report you can supply?
4. Are there guidelines on what you should aim for regarding the additional
statistics discussed on the standards, such as max, min, skew, kurtosis and
mean absolute?
5. Can you clarify what it means when you can state ‘tested to meet’ versus
‘produced to meet?’
6. Can you direct me to a document regarding planning and best practice
guidelines?
7. Do the vegetated area ground control points simply go on the bare ground
between vegetation?
8. How do you assess seamline mismatch?
9. If an orthophoto fails a column in Table B.3 (e.g. the RMSEr is okay, but the
accuracy at 95% CI exceeds the limit), do you select the accuracy class where
your project meets or exceeds all standards in a single row?
10. I note that many drone-based surveys seem to have a mean error much higher
than 25% of the RMSE. What does this information tell you about the quality of
the project, and how can you correct it?
11. I also noticed that nearly every drone software company reports accuracy as a
function of GSD, e.g. heights within three times the GSD. How are these related,
and is GSD really related to accuracy in any way? I did a project with a GSD of
1 cm, but I achieved 11 mm RMSE heights and mean of 3 mm.
12. Can you direct me to where I can read more about rigorous total propagated
uncertainty regarding photogrammetry?
James Wallace
University of Southern Queensland, Australia
Dr. Abdullah:
Due to the length of this list of good
questions, I will address them over the span of
several articles.
P
art
I
Question 1—
In section 7.8 of the stan-
dards, when it talks about ground control
accuracy being ½ RMSEmap, is RMSEmap
the desired/intended accuracy class?
Answer:
Yes, it is. The standards, in section
7.8 impose the following relationship between
the accuracy of the ground control points and
the derived product or map:
y
y
Accuracy of ground control designed for
planimetric data (orthoimagery and/or digi-
tal planimetric map) production
only
:
–
–
RMSE
x
or RMSE
y
= 1/4 * RMSE
x(Map)
or
RMSE
y(Map)
,
–
–
RMSE
z
= 1/2 * RMSE
x(Map)
or RMSE
y(Map)
y
y
Accuracy of ground control designed for
elevation data, or planimetric data and
elevation data production:
–
–
RMSEx, RMSEy or RMSEz = 1/4 *
RMSE
x(Map)
, RMSE
y(Map)
or
RMSE
z(DEM)
Therefore, if your photogrammetric processing
is solely for producing 2D planimetric data,
“The independent source of higher
accuracy for checkpoints shall be
at least three times more accurate
than the required accuracy of the
geospatial data set being tested.”
Photogrammetric Engineering &
Remote Sensing
Vol. 85, No. 6, June 2019, pp. 407–408.
0099-1112/19/407–408
© 2019 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.85.6.407
