CIPA 2003 XIX th International Symposium, 30 September - 04 October, 2003, Antalya, Turkey
529
NG OF
;e 27-29,
nent,
as terrestrial
1 advantages.
This statue,
The statue
of the whole
[ offering the
; during data
> parts. The
otos into one
ie 3D surface
igulation the
Finally, our
e recording
The statue
pproximate
The statue
pulled by
i the back-
3 described
istration of
cy analysis.
,iled struc-
ensions less
3 face of the
hnt of view
rple surface
is possible.
iokes of the
Due to the
be avoided
tremendous
reduce the
ta gaps and
l is devoted
camera Ko-
S-Z360 (cf.
on a Nikon
its (6.2 mil-
ize of 9/um.
i radiation,
red and the
of an inter-
3060 pixels
iurrounding
,ge distance
Figure 1: Photo of the statue Marc Anton which shows
the complexity of the sensed object.
to the sensed object of 5m, the average pixel size on the
object is 3mm. The laser scanner LMS-Z360 allows to cap
ture 360 ° scans with a vertical scanning range of max. 90 °.
The measurement rate of this system is 12000pts/sec. For
the range measurements a pulsed laser beam with a beam
divergence of 2mrad is used. The sensor provides a single
shot accuracy of ±12mm. For natural targets range mea
surements up to 200m are possible. The minimum angle
step width of this laser scanner is 0.01 ° offering an angu
lar resolution of 0.002 °. Additionally to the high frequent
range and angular measurements true color of the target’s
surface can be recorded by a one pixel camera. The time
of exposure for this one-pixel-camera is quite short due to
the high point measuring frequency resulting in an instable
radiometric quality.
With the help of these sensors 22 digital photos and scans
from 10 scanner positions (at every scanner position two
scans were acquired: one coarse panorama scan (360 ° with
an angle step width of 0.2 °) and one fine scan (angle step
width of 0.05 °) of the object) were captured (see Fig.
2). For the collection of the image data approximately
one hour was necessary, whereas for the 20 scans the data
acquisition time was approximately 2.5 hours. For the
observation of the higher object surfaces 5 photos were
taken from a ladder and two scanner positions were taken
using a higher scanner mounting.
For the identification of homologous points in the
photo and scanner data 29 signalized tie points were used
(see Fig. 2). 18 of these points were located on the statue
(quadratic retro-reflective foil with a size of 2cm by 2cm).
For the other tie points (11) retro-reflecting cylinders
(with a hight of approx. 7cm and a diameter of approx.
5cm) located around the statue were used. 10 minutes
per signal group were necessary to fix the retro-reflecting
tie points. For the selection of the most appropriate
position no complicated analysis is necessary. The aim
was to place the targets in a way that 4 of them can be
sensed from every sensor position in order to reduce the
amount of work for manual identification of un-signalized
tie points.
3 HYBRID ADJUSTMENT
For the generation of a surface model of the whole
sculpture it is necessary to transform all different sen
sor observations into one global statue co-ordinate system.
All observations in the scan data (distance measure
ments, horizontal and vertical angles) and in the image
data (pixel co-ordinates) can be described in a local
Figure 2: Scanner positions (red circles), photo posi
tions (blue rectangles) and signalized tie points (magenta
crosses). The three black triangles on the statue are used
to define the global statue co-ordinate system.
sensor xyz-co-ordinate system. For the description of
the observations in the global co-ordinate system the
transformation parameters of a spatial similarity transfor
mation of each sensor position must be derived in a hybrid
adjustment procedure considering individual weights for
certain observation types (cf. [Ullrich et al., 2003]).
Such a hybrid chained spatial similarity transforma
tion considering image as well as laser scanner data can
be performed with the photogrammetric software-package
ORPHEUS/Orient ([ORPHEUS/Orient, 2003]) on the
basis of homologous points. For this aim the measurement
of tie points in the image as well as in the scanner
data is necessary. For the adjustment of the data of
the statue Marc Anton the global co-ordinate system
was defined with the help of three control points in the
horizontal plane of the socket (cf. Fig. 2). To achieve an
unconstrained definition only 6 (defining the translation
and rotation of the global system) of the 9 co-ordinates
were fixed. For the definition of the scale of the global
system the range measurements provided by the laser
scanner were used.
The measurement of homologous points was per
formed manually in the digital photos as well as in the
intensity images provided by the laser scanner (see Fig.
3). All in all 33 tie points (29 signalized) were used. The
measurement of these tie points in the image data as
well in the coarse panorama and fine scans lead to 1300
observations for the adjustment. With the help of robust
estimation and data snooping 62 of these observations
(mainly from the measurements in the coarse panorama
scans) were eliminated so that in the final adjustment
1238 observations participated. The accuracies after the
adjustment were very satisfying: It was possible to de
termine the position of the laser scanner with an average
accuracy of 16mm in X-direction, 19mm in Y-direction
and 28mm in Z-direction (the definition of the coordinate
