International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
oriented, the whole process is fully automated. It is only
necessary to define the pixel size. For both data sets the pixel
size was set at 5mm. In total, four orthophotographs referred to
the roof (data set 1) and four orthophotographs referred to the
eastern facade (data set II) were produced. From those
orthophotos the best possible orthophoto-mosaics were
produced, and are given at Figures 4a and 5a.
While processing of the scanned data was performed using the
Cyclone software, this has no capability of producing
orthophotographs. So, the final TINs were exported to ASCII
files from Cyclone and these were imported to the Softplotter
software. With the new DSMs and the reinstated orientations,
new orthophotos were produced, with a pixel size of 5mm.
Finally, the corresponding orthophoto-mosaics were produced
for both data sets as illustrated in Figures 4b and 5b.
The visual examination of the final orthophoto-mosaics revealed
that:
e For dataset 1 (Figures 4a and 4b), there were no satisfactory
results for the left part of the last upper row of tiles (north-
eastern side of the roof) for both methods; the problem was
caused by the orientation angles of the images. Also, there
was a total failure in orthophoto production of the lower right
corner of the roof, because of the lack of photographic
coverage. For the remaining part, the orthophoto-mosaic
which was created by applying only photogrammetric
procedures gave results of better quality. The existing gaps on
the surface of the object due to occluded laser scanner data
influenced negatively the result, regardless of the density of
the data at the remaining area. It must be mentioned that the
results are the same using the laser scanner data with
decimation of 80% instead of decimation of 15%.
For dataset II (Figures 5a and 5b), the results of both methods
are satisfactory, except from some specific parts of the facade.
In particular, there were small areas on the right part of the
facade (at the lower level and a strip between the 2™ and the
4rd
3" level of the facade), where no photographic coverage
existed, and these are left blank in both final orthophoto-
mosaics. Also, there are some weaknesses at the top of the
dome, because all photo were taken from a lower level; and
also at the right half of the same dome. For this area the
results are better at the orhophoto-mosaic produced by using
data derived from laser scanning, as the geometry of
stereoscopic observation is very weak.
An accuracy control study was followed for both orthophoto-
mosaics of each dataset by applying a number of comparison
tests. It should be noted that all comparisons and checks were
performed using points from areas of the mosaics that had no
evident deformations.
Two comparative tests were made for dataset I:
i. the first test involved the evaluation. of systematic and
absolute errors of ten (10) premarked check points
distributed on the two mosaics. These check points were
different from the control points used during the
photogrammetric process. The mosaic-coordinates of the
selected points were checked against the corresponding
coordinates resulted by the surveying calculations.
ii. the second test involved the check of ten (10) selected
distances on the two orthophoto-mosaics. The end points of
each distance are clearly defined points on the mosaics. The
coordinates of those points are not measured by field
surveying techniques. The distance lengths vary from 0.16.
4.00 m, with random directions.
Table 1 gives the results of the two comparative tests and Figure
4c illustrates the results of the first test.
Test 1: Point errors (10 check points)
Photogrammetrically Mx = -2 mm My = 0 mm
produced orthophoto- Ox 729mm 6,-8mim
mosaic
Orthophoto-mosaic M, = | mm My = 3 mm
from laser scanner data | ox = 9 mm Coy = 3 mm
Test 2: Difference between distances (10 distances)
Mean of Differences Mg = 8 mm
RMS of difference RMS (dS) = 9 mm
Table 1. Evaluation of accuracy between the two orthophoto-
mosaics of data set I (north-western part)
o, - N(V;- My] /(n-1)
where: n - number of check points and
Vi - the difference between the 1 point coordinate from
the surveying estimation and the equivalent coordinate
from the orthophoto-mosaic (in both X, Y directions)
My, My 7 the mean of the V; differences in X and Y direction
M, = the mean of the differences in distances.
The analysis of the above statistical results gives that:
e There is no presence of systematic errors left in the final
orthophoto-mosaics either in the case of a . pure
photogrammetric procedure or in the case of laser scanning
data collection.
The values of the absolute deviations of the orthophoto-
mosaic using laser scanner data are within the accuracy limits
of the coordinates from the surveying estimation. On the
contrary, the absolute deviations of the photogrammetrically
produced orthophoto-mosaic are considerably larger in X
direction, and give final accuracy results acceptable for scales
only x 1:100.
e The differences between the two mosaics are small with a
total deviation less than 1 cm.
Two comparative tests were also made tor dataset 11:
i. the first test is similar to the first test made for dataset I. The
check was performed for 19 premarked check points and the
results are given at Table 2 and are illustrated in Figure 5c.
i. the second test involved the evaluation of the relative errors
of the two orthophoto-mosaics by using 26 check points,
whose coordinates had not been previously calculated by
surveying techniques.
The above statistical results indicate that:
e practically no systematic errors are detected at any of the two
mosaics
International
HL eal
(a) PI
(a)
