a)
wn
=
£1 + - - wows ded nilaupe-vilenionig we à À
0 | f |
0 2 5 10 15 20 30
Number of pairwise observations
b)
5
ws
TS
RMS error [cm]
N
2 5 10 15 20 30
Number of pairwise observations
RMS error [cm]
0 2 5 10 15 20 30
Number of pairwise observations
—9— 2 —#3 ——5 —— 10 — 15
Figure 5. Combining multiple and pairwise observations. RMS
errors in the check points when a varying number of multiple
observations was selected in each tie point area, as a function of
the number of added pairwise observations.
* Slightly better accuracy was achieved when combining
multiple and pairwise observations, than when using only
multiple observations.
4. CONCLUSIONS
In this paper, automatic tie point extraction has been discussed.
A new conceptual division of the tie point measurement process
into tasks is presented. The approach pays attention to the accu-
racy questions and the treatment of problems in the tie point
measurement.
The effect of the number, completeness and distribution of the
tie point observations was empirically investigated using the
342
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
OEEPE test block Forssa with 30 um pixel size. The investiga-
tion was not comprehensive, it was meant to give ideas about
the effect of some factors.
The factors affected especially the height accuracy. The accu-
racy of the block improved as the number of observations in-
creased. This effect was only valid up to a certain limit (15
points in each tie point area). The reason for this is either
limited accuracy of the method or more likely, that there were
no more significant observations. 3x3 and 5x5 distributions of
tie point areas were tested. The 5x5 distribution gave slightly
better results. The accuracy of the block deteriorated when the
completeness of the observations decreased, though the accu-
racy of the block using 4-neighbouring observations was nearly
as good as the accuracy using complete observations. À new
strategy for tie point measurement, carrying out both multiple
and pairwise matches, gave good results.
The RMS errors in the check points were in the best case, when
using automatic tie point observations: X: 1.8 cm, Y: 2.3 cm
and Z: 3.7 cm. The accuracy was on the same level as when
using accurate interactive measurements (X,Y: 2.3 cm and Z:
3.3 cm). The result with automatic measurements is very prom-
ising, because it seems that the accuracy can be further im-
proved as the knowledge about carrying out and handling the
observations increases.
Important topics for further investigations are the factors
affecting the accuracy and reliability of the block, like
distribution of and number of tie point observations, usage of
additional parameters and matching strategies (for instance,
combining multiple and partial matches). Investigations using
imageries of different scales and types are needed. It is also
important to incorporate intelligence in the systems, to increase
the level of automation.
References
Forstner W., 1995. Matching Strategies for Point Transfer, 45th
Photogrammetric Week, Stuttgart, pp. 173-183.
Honkavaara E., Hggholen A., 1995. A System For Automatic
Tie Point Measurement, Photogrammetric Journal fo Finland,
Vol. 14. No. 2, pp. 16-22.
Jaakkola J. and Sarjakoski T., 1994. OEEPE Research Project -
Aerotriangulation Using Digitized Images - Preliminary Results,
International Archives of Photogrammetry and Remote Sensing,
Vol. 30, Part 3/1, Munich, pp. 416-421.
Sarjakoski T., 1998. Automation in Photogrammetric Block
Adjustment Systems - On the Role of Heuristic Information and
Methods, Acta Polytechnica Scandinavica, Ci 88.
Schenk T., 1995. Zur automatischen Aerotriangulation,
Zeitschrift für Photogrammetrie und Fernerkundung, 63.
Jahrgang, 3/95, pp. 137-144.
Tsingas, V., 1992. Automatisierung der Punktübertragung in
der Aerotriangulation durch mehrfache digitale Zuordnung,
DGK, Reiche C, Heft 392, Munchen.
Tsingas, V., 1994. A Graph-Theoretical Approach for Multiple
Feature Matching and its Application on Digital Point Transfer,
International Archives of Photogrammetry and Remote Sensing,
Vol. 30, Part 3/2, Munich, pp. 865-871
KEY V
ABSTE
In ana
presem
rouglan
itl evide
fildrati
The re
initial
The te
high d
differe
of vie
particı
ihe gel
of gre:
their g
forms.
param
the de:
easine:
respec
consid
param
( relie
extremn
deriva
distan
terrai
1984 )
The re
the ge
aspect
accord
classif
topolo
geomo
macro
( 10*
( « 10
variat
this hi
The de
comp
from
under
simila
comps
perfor
presen
(rans:
