5.5 Effect of tie point type
In the block a large number of tie points ( in some cases
» 60 96 ) were signalised check points. To investigate the
behaviour of the block when only natural tie points are
used, additional measurements and calculations were
made in EPFL and FGI. In EPFL new block adjustments
were made so that all the check points were excluded,
leading at the same time to lower redundancy. In FIG all
the check point measurements in cases FIG 10-17 were
replaced by new measurements on natural points in the
vicinity of each check point, preserving the geometry of
the block. Check point coordinates were then calculated
after the block adjustments using spatial intersection.
The results of plane coordinates are presented in Figure
9. They show that in digital systems natural tie points are
as good as signalised points. The conclusion is the same
with Z-coordinate.
80
60
[EE .
mm 40 (89
gnat
pM
20... r
ml A ARR d
m m mm mn T
6e: m.m mom d
- 227 = 2
c2 e T
Figure 9. RMS values of plane coordinates in
measurements of EPFL and FGI, when all
possible (i.e. signalised and natural = sig)
and only natural (nat) tie points have been
used. The blocks have been adjusted using
dense control and additional parameters.
5.6 Theoretical accuracy estimates
The accuracy of check point measurements on the
images has direct influence on the accuracy estimates of
ground points. To avoid this problem, also theoretical
estimates, calculated in the block adjustments, were
analysed. However, the analysis of the estimates of the
elements of exterior orientation showed that these values
alone do not give any reliable information for
comparisons. The large differences in block geometry
were the main reason.
6. CONCLUSIONS
This project on digital aerial triangulation was imple-
mented as a multi-site comparative test project. Nearly all
of the triangulations carried out by the participants would
satisfy normal requirements of any mapping or data col-
lection project. This is a very positive outcome, especially
when the large variety of the methods and systems and
also the experimental nature of these systems is taken
into account. The main results of this test project are
related to accuracy. The following detailed conclusions
can be made:
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
e Digital aerial imagery seems to have the potential for
accurate aerotriangulation. In the best realisations,
the accuracy is at the same level with the most
accurate conventional methods. With the proper
usage of automation, which means a high amount of
tie points, high quality coordinate measurements on
the image, an even distribution of tie points, and good
ties between image strips, the accuracy potential of
digital imagery exceeds that of analogue data and
conventional methods.
e In this test, pixel size was, in most of the cases, a less
important factor than expected. It was in the most
accurate methods, almost neglectable. On the other
hand, pixel size is, in systems using feature based
matching for tie point measurements, the most critical
factor.
e A measuring resolution of half a pixel is sufficient also
with small amounts of observations, if the pixel size is
15 um.
e |n methods using area based matching, signalised tie
points are not significantly better than natural details.
e Planimetric accuracy is in completely visual methods
using digital imagery, in the best cases, at the same
level with more advanced methods. The heights are in
most cases slightly poorer.
e The relation of height to plane accuracy is smaller
when area based image matching has been used on
tie point measurements than in visual methods.
e The used theoretical accuracy estimates give too
large values for height accuracy. The reason is the
incorrect stochastic model used in block adjustments.
e Sensitiveness to sparse ground control is, when using
digital methods, very often larger than with
conventional analogue methods.
The conclusions above have been drawn from the results
and experience obtained in this case. It is especially
characterized by the use of large scale photography in
urban area. The results could be different in other condi-
tions and with different project parameters. The most
important parameters are the scale of the photography,
the block size, the shape and size of the signals, and the
use of in-flight GPS measurements. In these respects,
there are many opportunities for further development and
research.
References from Other Literature:
Jaakkola,J., Sarjakoski,T. 1996. Experimental Test on
Digital Aerial Triangulation. OEEPE Official Publication
No 31.
354
KEY W
ABSTF
The ar
resolve
applica
drift pe
system
poor C
constel
new ap
The sh
reason
the ne
had cy
additio
The in
and bi
numbe
operati
for hei
determ
even |
phases
The ce
kineme
have .
accura
depen
and th
over di
decime
measu
precise
slips c
correct
systerr
The ge
uses |
adjust
project
adjustr
GPS-p
they al
the tur
the co
change
