International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
Figure 13: — Aerial image in the suburb of Brussels
(©EUROSENSE).The scene is 270 x 340 m? and the im-
age has ground resolution of 8 cm.
The 3D reconstruction is performed using robust parameter es-
timation in order to limit the influence of image segmentation
defaults and of the presence of outliers: it is with this aim in view
that the RANSAC approach has been selected. The criterion to
evaluate the quality of a relative registration of both data is then
the proportion of outliers revealed by the RANSAC procedure.
Because of the global convexity of this criterion a simple Nelder-
Mead simplex method may be used to recover the deformation
between the aerial image and the cloud of laser points. The ex-
periments show that rigid deformations can be recovered with this
approach.
Further experiments should be undertaken in order to prove the
efficiency of the approach for the calibration of airborne laser
data:
e application to other data with different characteristics: im-
age resolution, laser scanning systems (oscillating or rotat-
ing mirror, optic fibers, ...), laser point density, etc.,
® test for more complex deformations of the laser points cloud:
composition of translation, rotation and curvature; and de-
formations with higher frequencies.
e integration in a global laser point calibration system, espe-
cially verify that the precision reachable using this registra-
tion approach is sufficient for calibration purposes,
With a different objective than the registration of a laser point
cloud and an aerial image, the tools presented in this study may
also be of great interest for a fine and detailed 3D reconstruction
of a urban scene using 3D laser points and several aerial images.
Acknowledgment: The author is grateful to EUROSENSE for
providing the aerial images and the laser points, and to Cyril
Minoux for his recursive implementation of the parameter space
scanning algorithm.
Figure 14: Footprint of the laser points (EUROSENSE). The
elevation is ranging from 11m to 45m.
REFERENCES
Baltsavias, E., 1999. Airborne laser scanning: basic relations
and formulas. ISPRS Journal of Photogrammetry and Remote
Sensing 54, pp. 199—214.
Bretar, F., Pierrot-Deseilligny, M. and Roux, M., 2003. Estimat-
ing intrinsic accuracy of airborne laser data with local 3d-offsets.
In: Workshop ISPRS on "3-D reconstruction from airborne laser-
scanner and InSar data", Vol. XXXIV, PART 3/W13, Dresden,
Germany.
Fischler, M. A. and Bolles, R. C., 1981. Random sample con-
sensus: A paradigm for model fitting with applications to im-
age analysis and automated cartography. Communications of the
ACM 24, pp. 381-395.
Nelder, J. and Mead, R., 1965. A simplex method for function
minimization. Computer Journal 7, pp. 308-313.
Rousseeuw, P. and Leroy, A., 1987. Robust Regression and Out-
lier Detection. John Wiley & Sons, New York.
Roux, M., Maitre, H. and Girard, S., 1997. A step towards stereo
reconstruction of urban aerial images. In: Joint ISPRS Commis-
sion III/IV Workshop on 3D Reconstruction and Modelling of
Topographic Objects, Vol. 32, part 3-4W2, Stuttgart (Germany).
pp. 107-114.
Schenk, T., 2001. Modeling and analyzing systematic errors in
airborne laser scanners. Technical report, The Ohio State Univer-
sity. Technical Notes in Photogrammetry No 19.
Suk, M. and Chung, S. M., 1983. A new image segmenta-
tion technique based on partition mode test. Pattern Recognition
16(5), pp. 469-480.
Zhang, Z., 1997. Parameter estimation technique: a tutorial with
application to conic fitting. Image Vision Computing 15, pp. 59-
76.
KEY
ABS]
Mixec
to ent
spatia
user’s
reactk
This €
will p:
crucia
and c«
togeth
visual
to illu
repres
by apr
manag
The "I
1) defi
the rez
or aug
real w
erated
is in c
totally
though
it is tl
"true"
If augr
represe
overlay
mental
pen in
sion ke
ments.
With ai
spatial
and me
tion an
scene. |
plays, t