LASER SCANNER
r, alvaroml@ufpr.br
anada —
estrial surface tends to
jlity for mathematical
1 the laser scanner, has
>d to obtainment of the
; applied to obtainment
cost in the productive
gy used and the results
btained from low cost
s were provided by an
CCD dimension of the
cel resolution in terrain
1000,00 meters of the
is OPTECH/INS GPS
er presents the height
without height features
was developed in the
exterior orientation of
ts of the entities were
conventional 3D stereo
: of 1/6000). The paper
face (See Ackermann
the development of
n information about
atures, less qualified
he productive process,
of methodology and
ctor tri-dimensional
y 3D Monorestitution
ed with small-format
scanning system. A
i covering performed
DSC-F717 model, 52
0 Pixels, was utilized.
an approximate flying
xel with approximate
+
de
inated from a survey
flying height of 1000
° 0.5 meters, obtained
stem, belonging to the
~LACTEC-UFPR
\
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
An aerial dataset carried out with a WILD RC-10 conventional
camera in the scale of 1/6000 was also employed, in order to
obtain control points and verify the accuracy obtained in the
proposed methodology.
2. TEST AREA
A digital image covering the region of the Centro Politécnico of
the Federal University of Paraná was utilized in this research
(See picture 01). This image is part of the aerial covering
carried out by AGRITEC Agrimensura e Aerofotogrametria
S.A. in July 2003. A small-format Sony DSC-F 717 digital
camera was employed, with resolution of 5.2 Mbytes, CCD of
2520 x 1960 pixels. The aerial covering was done in a flying
height of approximately 750 meters, resulting in pixel with
spatial resolution on the terrain of 25 centimetres.
Taking into account the characteristics of the buildings, one can
divide the image into two distinct regions. The first, on the left
side, has part of a residential neighborhood where the buildings
are smaller and close to one another. The second, in the central
part and right side, has bigger buildings and more space
separating them. This sub-division of the image into two main
regions will be discussed in the next chapters. It was done
simply to analyze the accuracy obtained in the experiment
carried out.
Figure 1: Aerial image taken with Sony camera employed in
this research.
3. GEOMETRICAL PARAMETERS OF THE DIGITAL
CAMERA
The parameters of interior orientation of the Sony DSC-F717
digital camera were determined with three calibrations utilizing
information of three different dates. The same operational
conditions and techniques were employed for the taking of
pictures of the test field existing at Federal University of Paraná
- Centro Politécnico. Stability cónditions were analyzed to
determine these parameters for a certain period of time in the
utilization of the camera. The values employed in this research
are presented in table 01. Other information about calibrations
performed and stability of the parameters of interior orientation
may be found in MACHADO et al., 2003, HABIB et al., 2003
and HABIB et al., 2002.
Interior Orientation Parameters
Parameters Determined values
e 10.082 mm
Xo -0.253 mm
yo -0.151 mm
k, -2.18915186e-03 mm”
k, 2.75934941e-05 mm?
ks 0.000
P -1.47185370e-04 mm”
P, -1.42394475e-05 mm”
Table 1: Geometric parameters of the camera.
4. LASER SCANNER DATA
A laser scanner survey of Centro Politécnico region covering an
area of approximately 7.2 km2 was conducted in flying height
of 1000 meters, with resolution of approximately 0.5 meters,
obtained through OPTECH/INS GPS APPLANIX System,
property of the Institute of Technology for Development —
LACTEC — UFPR.
This work utilized a cloud of points with coordinates X, Y, Z,
originated from the last echo. Only the area covered by the
digital image was delimited, resulting in a group of
approximately 800,000 points. The processing of these points
was carried out with the purpose of selecting only those
identifying the borders of edifications contained in the image.
The mathematical algorithm developed performed the
processing in two stages. In the first, all the points belonging to
a scanning laser strip were filtered utilizing the technique of
simplification of polygonal based on the algorithm of Douglas-
Peucker, as shown in DOUGLAS 7 PEUCKER, 1973. In the
second, each of the points obtained in a profile was analyzed
with the purpose of selecting only those that presented a height
difference bigger than a threshold furnished, relating to the
former or latter point, as presented in AXELSSON, 1998.
Figure 2: Shows the results obtained with the application of the
methodology suggested in region of big constructions. One can
verify that the mathematical algorithm proposed managed to
evidence satisfactorily the laser points existing on the borders
of main buildings. Therefore, it has the disadvantage of not
excluding defining border points of the main trees existing in
the image.
Picture 03 shows the results of application of methodology in
the small buildings residential area. In this case, one can see
the low performance of the algorithm proposed, for due to the
little space separating one building from the other and the laser
system resolution, the algorithm fails to present some borders
points desired. On the other hand, by diminishing the
threshold, other borders not desired are selected.
