2004
M
. The
osen
It any
'M as
n for
Arrow
soft
'M of
anted
nes.
(left)
lines
ed by
(right)
in the
1ethod
In the
au, the
should
s such
s and
of the
Itering
f sand
It js,
re is à
objects
International Archives of tl
Shaded view of a ter filtering (Stockerau);
b: buildings that could not be eliminated.
Figure 8.
Figure 9. Shaded view of a DSM b (1
filtering (right) from the Schee: st site.
The Schneealm test site was selec whether ou
filtering method could be used
wooded areas. Given the image scale of tha te, the widtl
M:
1
of the elevation grid delivered by MATCE
10 m. Consequently, the parameters had to a differe
way than described in section 4.2, even m » because only
very few terrain points could be expect | The
out 3 hos: ) be 100 m
hosen ratheı
grid width for the first thinning-
arameters for the weight functio
ictive in order to eliminate as ma iin points as
2 (A s =; = (03m. The ten odelled very
ly after the first loop of thinning
d by a DTM with a grid wid
iteration, the tolerance band was only points
above the initial DTM were all points
below that DTM were regarded 1 in points. The
| hinned out
ith of 30 m.
5noa prediction was applied using restrictive
of the weight function
th = s = = 0.25m), again in order to only eliminate points
| failed to eliminate the
ring the data,
1 the second
original points classified as
est point within
using the 1ow
Hh
parameters on the positive branch
However, the
it-te i this stage, so that the ' intermediate
surface model was not close enough to the ‘terrain.
C third iteration could not succeed, either. Our
ts because too
g method could not deliv ;
lew terrain points were provided by image matching, so the
algorithm was not able to eliminate the influence of the off-
rra n the trees (figure 9). The drawback of digital
Image matching methods compare in wooded areas
Was ob viol
Photogrammetry, Remote Sensing ai
Information Sciences, Vol XXXV, Part B- tanbul 2004
)oint clouds into un points
in order & ca IM
data or image matching. We have sl lOWw à
method originally designed for the filtering of ALS « in b
generate a DTM from the DSM created by imag
chniques. The basic difference LS point
ribution of the
matchin
clouds and the image matching results is the
points in wooded and densely developed urban areas. The
sequence of the applied strategy in SCOI ught be the same
to be adapted to the
a sets, but the parameters ha
hi
ics of Ms from image £
m image matching techniques does still contain off-
{
ts in spite of the filter methods integrated he
he Is concerning
tha et] gives acceptable results for urb: The
influence of off -points 1s widel
at the cost of
cffects can be eliminated by the inclusion of break |
ver, these smoothing
thing. How
so that
ed. The
results IC VOOG as are not satisiving because terrain
a very good re entation of the terrain ca
points are acquit natching techniques. ALS dat
better sui ation in forested areas
Pfeifer, 1998
when usi ter riginal data from the matching
ement of the results is expected
process or if the d of othing in the matching process 1s
possible in order to eliminate
selected t
undesirable pre- ects.
KNOWLEDGEMENTS
This work w pported by the Austrian Science Foundation
(FWF) under pr: 15789 and by the Australian Research
( overy Project DP0344678
Council (ARC) under Di
REFERENCES
Pfeifer, N., Dorninger, P., 2002. Application of the
isl erpo for DTM determination. In: ZAPRSIS
x \ ST.
'ugung digitaler. Gelàndemodelle durch
isch Bik rdnung. PhD thesis, Institute of
rsity of Stuttgart. DGK-C 418.
; 1998. Determination of terrain models in
wooded : "n erial laser scanner data. ISPRS J. of Ph. &
DC <A
Kraus, 000. Photogrammetric Band 3. Topographische
Inform ssysteme. 1" ed., Dümmler Verlag, Bonn.
Kr KP. . Generation of Digital Elevation Models. In:
© Course in Digital Photogrammetry, Institute of
PI etry, 5onn University, and Landesvermessungsamt
Nordrhein-Westfalcn.
lution, 2001. Summit Evolution, Digital
or use with Windows NT or Windows 2000.
wal. DAT/EM Systems International.
2 Web page: www.inpho.de, 15.3.2003;
