he
is
he
be
ity
ior
OW
int
ry
lot
re
ler
he
le
nt.
ue
ng
int
all
19
ill
19
of
1e
oS
al
al
4. EXPERIMENTAL TEST
The purpose of the practical work was to test the state of
the art of the digital image mapping model discussed
before, and it meant to analyze the capability and
possibility of applying the automatic DTM generation
methods from stereopairs of SPOT imagery intended for
1:50 000 scale image maps.The experiment was an
automatic DEM generation, performed on a stereopair of
SPOT imagery (panchromatic, level 1A) covering a hilly
area geographically located in the Aix-en-Provence region
in the south of France.
Due to some limitations caused by imperfection of devices
and lack of accurate knowledge about parameters involved
in automatic DEM generation function it is still doubtful
whether this method will satisfy all necessary height
information specifications required for image mapping at
scale 1:50,000.
However, regarding the scale of the image map e.g.
1:50,000 and a contour interval of about 10 to 20 m (for arid
area), it can be concluded that the final image map must
have the planimetric accuracy about 10 to 15 m (i.e. (0.2 to
0.3)*50,000/1000- 10 to 15 m) and altimetric accuracy
(contours) about 3 to 7 m (i.e. 0.3 * (10 to 20) 2 3 to 7 m).
Therefore, the question to be answered is " Does the
automatic DEM generation method provide height
information, accurate enough for both, differential
rectification of image, reaching to the 10 to 15 m planimetric
accuracy, and also contour line generation. To answer this
question the SPOT panchromatic stereopair (level 1A) and
the infomap 3142 at scale 1:25,000 provide the primary
information. The processing is applied to an image dataset
of 20000*10000 elements 200 km? on the ground. The
10*10 m grid automatic DEM is generated, and will be used
to correct the image data, and be compared with DEM
generated from contours.
The 20 m interval manual height measurement on
TRASTER T10 analogue stereo plotter along three different
lines within the area should be accomplished to obtain three
different profiles. This is done to find out the systematic
error on automatic generated DEM. The 10 m interval
contour lines of the 1:25,000 scale topoplot will be digitized
to produce Digital Elevation Model. The DEM will be used:
to evaluate the relative accuracy of automatic generated
DEM.
4.1 Quality assessment of DEM generated automatically
The purpose of this section first is to evaluate the quality of
automatic generated DEM by analysis of statistical result of
residuals among two DEM, Automatic one and reference
DEM which has been generated by digitizing contours. The
Second purpose is to evaluate the planimetric differences
between two orthoimages rectified from left and right
stereopair with automatic generated DEM. In fact, There
should be no planimetric differences between same objects
being found in two orthoimages from left and right, if the
DEM applied in rectification had been accurate enough.
777
4.2 Altimetric accuracy evaluation
To evaluate the two DEM, 1500 points along three different
lines with interval 20 meters had been selected and the
differences in height between corresponding points in each
line were calculated.
Subsequently, the results of this computation are as follows:
Statistical result of comparing 3 profile on both DEM
Mean value of Standard
residuals deviation of the
residuals
Profile 1 6.5 8.5m
profile 2 8.2 8.7m
Profile 3 3.2 10.2 m
Table 8.2 Statistical results of differences on three profile
derived from DEM generated automatically and DEM
generated from digitizing contours
According to (Imhof, 1982) , it is obvious that the all
differences have same direction therefore a systematic error
exists. To find out the sources of this systematic errors. It
was decided to measure the same profile by manual
method with TRASTER T10 just by putting the floating mark
on the ground on the same points that we had the height
from two existed DEM, and then compute the statistical
results for profiles driven from contours and from manual
method.
Subsequently, the results of this computation are as follows:
Statistical result of comparing 3 profile
Mean value of | Standard deviation
residuals of the residuals
Profile 1 3.02 m 8.8m
Profile 2 -0.35 m 9.04 m
Profile 3 -0.60 m 8.3m
Table 4.2 Statistical results of differences on three profile
measured directly on stereo plotter and Three derived from
DEM generated from digitizing contours
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
