International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
Figure 2: Tie surfaces for adjustment of the parallel strips in
the Eelde block
by boolean operations to find the relevant overlapping areas
that were not segmented so far. Association of counterpart
points (i.e., from other strips) to a segment was performed as
follows, for each counterpart strip points that are in or close
to a segment were collected and a plane was fitted through
them, points that exceeded a preset tolerance were rejected.
This procedure introduced a safeguard to exclude erroneous
points to enter the adjustment from the outset. Result of
the extracted tie surfaces for the an adjustment of only one
sub-block are illustrated in Figure 2. These segments serve
and the tie surfaces for the adjustment.
The results of the adjustment of the whole block are sum-
marized in Table 1, the variances of the estimated offsets are
very small and are therefore not listed. As Table 1 shows the
magnitude of the planimetric offsets is bigger than the height
offsets by an order of magnitude. The positional offsets reach
the order of tens of centimeters whereas the height offsets are
on the order of a few centimeters, the biggest among them is
four centimeters only. These results indicate that a 1D adjust-
ment of the data is insufficient for the adjustment of airborne
laser data. The relatively small height offset can be attributed
to the higher level of accuracy in the height determination,
or to a 1D adjustment of the data that was performed be-
fore the data was delivered. Figure 3 shows the offsets prior
to the adjustment and the results after the dataset was cor-
rected for the offsets. As can be seen there are noticeable
planimetric offsets before the adjustment whereas the offsets
in height can hardly be noticed. The offsets were eliminated
after the adjustment parameters were introduced. The re-
sults of the adjustment were also evaluated by a comparison
of the fitting accuracy of a surface to a segment (that consist
of points from more than one strip) before and after the ad-
justment. The results show an impressive improvement. For
tilted surfaces where offsets are noticeable the fitting accu-
racy for surface that was reduced from about 35cm before
adjustment (where the fitting accuracy of a segment from
one strip only was 5cm) to a fitting accuracy of about 6cm
after adjustment. Post adjustment results show indeed that
the positional offsets were eliminated.
Table 1 shows that the offsets within a sub-block are more or
less of the same order, however they are not the same. Fig-
ure 4 that shows the variation in the magnitude of the offsets
[ 1D. J 5Xo[m] | dYo[m] | 9Ze[m] ]]
PL 0.05 -0.01 0.04
P2 -0.17 -0.30 0.02
P3 -0.24 0.01 0.02
P4 -0.08 -0.25 0.02
P5 -0.16 -0.11 0.01
P6 -0.11 -0.21 0.04
P7 -0.25 -0.06 0.01
P8 -0.11 -0.14 0.00
X9 -0.27 0.29 0.01
X1 -0.46 -0.33 0.02
X2 -0.43 -0.49 0.01
X3 -0.56 -0.21 0.02
X4 -0.20 -0.31 0.01
X5 -0.48 -0.07 0.02
X6 -0.33 -0.11 0.02
XT -0.34 -0.02 0.01
X8 -0.18 -0.11 0.01
X9 -0.21 -0.37 0.03
X10 -0.34 0.28 0.01
Table 1: Offsets between for the individual strips. Strips
from the parallel sub-block are denoted by P, strips from the
crossing sub-block are denoted by X.
(the offsets norm) for the cross strips sub-block and indicates
this also graphically. It is therefore not advisable to consider
the offsets constant for a whole block. To test for variations
of the offsets within a strip the offsets were computed for
smaller sections within the strip but did not reveal significant
changes, therefore the strip unit seems to suit here. A more
detailed inspection of a few horizontal surfaces have indicated
that there are some trends that appear to arrive from angular
biases (such as mounting or INS biases). The magnitude of
these trends is of smaller order but still require treatment.
Extension of this work will concern with their elimination.
5 Concluding remarks
Reaching the potential accuracy of laser data and eliminating
artifacts requires the removal of systematic errors from the
data. A strip adjustment formulation enables removing both
errors that were not properly eliminated before takeoff and
ones that occurred during the mission. By using a system
driven solution in the current modeling the actual errors in
the system can be removed. The proposed model offers a
natural way for eliminating the systematic errors as it con-
strains the laser points to the surface. The selection of a
surface based model enables using general topography and
natural and man-made surfaces for the adjustment and do
not require distinct object in the overflown region. Results
of applying the model on a block consisting 20 strips among
which ten strips were taken in a “cross-strip” pattern have
demonstrated the existence of significant positional offsets in
the data of one order of magnitude bigger than the ones in
height.
REFERENCES
Bretar, F., Pierrot-Deseilligny, M., Roux, M., 2003. Estimat-
ing image accuracy of airborne laser data with local 3D-
offsets, International Archives of Photogrammetry and Re-
Internat
pds sd a Ka
*
Figure 3
adjustme
Magnitude [m]
Figure 4;
Cross stri
