PE&RS November 2015 - page 832

832
November 2015
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
to Table B5 of the new standard, which provides users with
guidelines that can be followed during the transition period
from the old standard to the new. Table 2, which is adopted
from Table B5 of the new standard, lists the recommended
accuracy figures that can be assigned to orthoimagery based
on its pixel ground resolution, in this case 7.5 cm. Bear in
mind that both tables, B.5 and 2, are based on the practical
experience and knowledge we gained over the last decade
and a half using medium and large-format metric cameras
and it
should not
be used for stating products’ accuracy
from non-metric camera products. In addition, such gained
experience using medium- and large-format metric cameras
is mainly derived from imagery with GSD of 7.5 cm (3 in)
GSD or larger. Imagery with GSD of around 3 cm is rarely
acquired for mapping purposes using these metric cameras.
Stringent ground control and airborne GPS accuracies are
required when dealing with very high-resolution imagery, i.e.
GSD of 3 cm or smaller, if you expect to meet the “2 x Pixel
Size” accuracy criteria using metric cameras.
Table 2 Recommended imagery resolution and products accuracy.
Common Orthoimagery
Pixel Sizes
Recommended Horizontal Accuracy Class
RMSE
x
and RMSE
y
(cm)
Orthoimage RMSE
x
and
RMSE
y
(pixels)
Recommended use
7.5 cm
≤7.5
≤1
Highest-accuracy work
15
2
Standard mapping and GIS work
≥22.5
≥3
Visualization and less-accurate work
Table B5 of the new standard and Table 1 in this answer
suggest that the current stage of technology surrounding the
mapping process can support the production of orthoimagery
accurately up to one-pixel. This is assuming that the highest-
quality production practices are followed. Although it may
be possible to achieve, I do not advise the use of the one-
“The new standard merely defines the
horizontal accuracy class of a product
by the root-mean-square error (RMSE)
required for that product”
pixel accuracy as an accuracy measure on regular bases. I
believe that during this transition period and during this
stage of mapping technologies, a measure of accuracy that
is equal to 1.5 times the pixel resolution should be adopted
when necessary to represent the highest current product
accuracy derived from metric cameras. Therefore, I suggest
using the following relationship between pixel resolution
and map accuracy for the time being until users empress the
new standard by defining map accuracy independently from
imagery ground resolution:
Horizontal Accuracy of Map or Ortho is RMSE
x
or
RMSE
y
= 1.5 x Pixel Size (or GSD)
“During the past two decades and
since we transitioned to softcopy
photogrammetry using either scanned
film or digital camera products, we
adopted a common practice to relate
the imagery ground pixel resolution to
the accuracy figure that is assigned to
a map scale according to the legacy
ASPRS map accuracy standard of 1990”
The 1.5 times the pixel resolution accuracy measure is
a middle ground between what we practiced during the
legacy standard (i.e. “2 x Pixel Size”) and what is currently
possible from the mapping process (“1 x Pixel Size”). Future
improvements in sensor technologies and the mapping
process may alter such measure, and it may lead us toward
one-pixel accuracy measure or better.
**Dr. Abdullah is Senior Geospatial Scientist and Associate at
Woolpert, Inc. He is the 2010 recipient of the ASPRS Photogrammet-
ric (Fairchild) Award.
The contents of this column reflect the views of the author,
who is responsible for the facts and accuracy of the data pre-
sented herein. The contents do not necessarily reflect the offi-
cial views or policies of the American Society for Photogram-
metry and Remote Sensing and/or Woolpert, Inc.
“Bear in mind that both tables, B.5 and
2, are based on the practical experience
and knowledge we gained over the last
decade and a half using medium- and
large-format metric cameras and it
should not be used for stating products’
accuracy from non-metric camera
products”
819...,822,823,824,825,826,827,828,829,830,831 833,834,835,836,837,838,839,840,841,842,...882
Powered by FlippingBook