ul 2004
atching
ifferent
ult. in
"rtunity
and
pacing
fferent
GPS
inually
ure 7)
"menos
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
Figure 8 and Table 1 show the comparisons between each
automatic DTM (correlated points) and a manually measured
one. In both the b) and c) cases, using seed DTMs does not
improve substantially image matching results. Similar
conclusions were obtained for the other test areas.
In our case of study. indeed, it was possible solving DTM
mistakes only by a post editing process. In particular, in the
neighbourhood of structures, top-and-base breaklines were
introduced; instead, in some sandy arcas, manual editing was
conducted point by point.
274024 274026 274028 274030 274032 274034 274036 274038
AO +
seh ++
3269594 + ;
3269592
3269590 -
3269588 | x
3269586
3269584 | #
32695824
3269580
3269578 | + +
dt.
3269576
3269574
2) + 06% + 562% + 259% + 179%
274024 274026 274028 274030 274032 274034 274036 274038
* *
EROR CER T .
32695964 + + .-Ft RT
+ + + s
.+ + +
+ +
32695044
| fod
3269592
3269590 *
|
3269588
3269586
3269584 | +
3269582 -
3269580
3200578] +}
T 7
3269574 | T oed
b) 4 03% + 558% + 239% + 20%
al
274024 274026 274028 274030 274032 274034 274036 274038
Mp > +
+de +
3269596 + 4 *
3269594 '
3269590
3269588
3269586
3269584
3269582
3269580
3269578- 4
3269576
3269574
e 4 +
oF 09% 4 569% + 247% + 175%
+ La
- -2 to -0.3 4» -0.3 t0 0.3. 4 0.3 to 0.6 + 0.6 to 3.5
Figure 8. Test area: classed plot of residuals (m) between
automatic and manual DTMs. Black dots represent seed points.
a) b) C)
Minimum -1.89 -0.54 -0.94
Maximum 3.04 3.00 2.53
Mean 0.34 0.36 0.33
Standard deviation 0.39 0.37 0.35
Table 1. Test area: statistical summary of residuals between
automatic and manual DTMs (values in m).
Finally, we compared the elevation data coming from kinematic
GPS survey with those deriving from photogrammetric DTM.
As in above reported tests, +30 cm was considered the
maximum tolerable value for the differences between GPS and
photogrammetric elevation data. That limit comes from
considering the image scale, the mean stereoscopic base of
about 40 m and the pixel size.
Observing figure 9, showing a portion of the comparison, it can
be noted that the higher inconsistencies are localized around
West temenos wall. This trend is explainable considering that
this area is covered only by two images with a little stereoscopic
base (about 10 m): in such condition we could accept up to +1m
differences. Such a problem stresses that small format low-
height photogrammetric systems do not allow for an accurate
control of the base and of the flight height.
As for the rest, however, differences are acceptable except for
really localized problems coming from structures or shadows.
The DTMs study has therefore showed the effectiveness of ATE
module in producing digital surfaces; it must observed that the
automatic procedure can give problems when the aim is the
contour representation, because image matching extracts points
without an interpretation of object morphology.