336
May 2019
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
For Topographic Mapping Products
According to the new ASPRS positional
accuracy standards, for a digital surface model
or digital terrain model to be produced to meet
a vertical accuracy class of 1 cm, the following
conditions needs to be satisfied:
A. The ground control used for the aerial
triangulation process needs to be surveyed
to a horizontal accuracy of 0.25 cm as
RMSE and vertical accuracy of 0.25 cm as
RMSE.
B. The accuracy of the aerial triangulation
needs to be within 0.50 cm as RMSE
horizontally and 0.50 cm vertically as
RMSE.
Looking at the stringent horizontal and vertical
accuracy requirements for the ground control
survey of 0.25 cm, an experienced mapper can
easily realize that real-time kinematic-GPS
field surveying practices cannot be utilized
for this type of survey work and that costly
differential levelling is the only method that
can achieve such accuracy. The high cost of
this highly accurate ground control survey is
prohibitively expensive and not suitable for low-
cost UAS-based products. The new UAS-based
mappers need to understand that most of the
field survey work conducted to support aerial
mapping usually meets an accuracy of 2 cm (as
RMSE) unless otherwise requested. Ground
control points surveyed to an accuracy of 2 cm
are only useful to produce mapping products
that are accurate to 8 cm according to ASPRS
standards and it does not support the 1-cm
accuracy claim that many UAS-based mappers
advocate.
Summary
Finally, our success in using UAS for mapping
product generation can be credited to the past
achievements within the photogrammetric
community and the introduction of several
innovative approaches by the non-mapping
community. This collaboration between the
two communities has resulted in an extremely
efficient workflow for processing UAS products.
Without this cooperation, the use of UAS for
mapping would not be as prevalent as it is
today.
“The SfM method
solves the camera
positions and
scene geometry
simultaneously,
using a bundle
adjustment of
highly redundant
measurements
based on matching
features in multiple
overlapping images
acquired from
different locations.”
they claim their products are produced to meet
an accuracy of 1 cm or better. This claim is
problematic for the following reasons:
Wrong Approach for Accuracy Evaluation
Product accuracy according to ASPRS standards
must satisfy the following conditions:
A. Accuracy should be assessed by using a
set of independent surveyed (or derived
from other sources) ground check points
that are more accurate than the tested
products. Independent check points are
ground control points that are not used in
the aerial triangulation for that project
B. To make it a valid statistical sample, there
should be at least 20 check points used in
the accuracy assessment regardless of the
project size.
C. Check points should be at least three times
more accurate than the tested product.
Besides violating the ASPRS standards
conditions on accuracy evaluation, many
of those providers report the results of the
ground control fit in aerial triangulation to
express their products accuracy. This is the
wrong approach as the aerial triangulation
subjects the ground control points to a
stringent constraint, which disqualifies it as an
independent check point.
For Planimetric Mapping Products
According to the new ASPRS positional
accuracy standards, for an orthorectified image
or planimetric vector mapping product to be
produced to meet a horizontal accuracy of 1 cm,
the following conditions need to be satisfied:
A. The ground control used for the aerial
triangulation process needs to be surveyed
to a horizontal accuracy of 0.25 cm as Root
Mean Square Error (RMSE) and vertical
accuracy of 0.50 cm as RMSE.
B. The accuracy of the aerial triangulation
needs to be within 0.50 cm as RMSE
horizontally and 1.0 cm vertically as
RMSE.
