nbul 2004
the graph.
chosen by
ical model
ities, such
| may still
2) — how
f previous
ot) knows
e root. By
s and their
more than
que branch
new iden-
r voxels in
f the scene
thm needs
t will find
it root. All
ling has to
Connected
ation algo-
he bottom
e an unas-
farted with
voxels in-
r that tree.
ned voxels
ed compo-
cing skele-
ponent La-
ly. Subse-
| sets, such
an be seg-
> sorted ac-
to different
ponents do
s of noise.
omponents
n a certain
y trees and
th a unique
| voxels ac-
| by raster
transform
algorithm,
he problem
cords) of a
dimension.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
Usually this is a feature space, and the purpose is to find those À
records in a database that are most similar to a query specifica-
tion; the number of considered attributes specifies the dimension-
ality.
In our case the space is Euclidian with just 3 dimensions. The
skeleton provides the data base, and we need only 1 neighbor
(skeleton voxel) in each query (tree voxel). In this way all tree
voxels will be labeled (Fig. 7).
Figure 7: Segmentation. Left: segmented skeleton; right: seg-
mented voxel space
3.8 Point cloud segmentation
Finally, we use once again the routine that originally transferred
all the laser points into the voxel space. This time the purpose is
to see into which voxel each laser point would be transferred, in
order to assign the voxel's label to the point (Fig. 8).
4 CONCLUSION
In a rather straightforward fashion we transferred a carefully cho-
sen selection of basic and advanced 2D raster (image) processing
algorithms into the 3D domain. The result is a flexible set of
tools, which we applied to segment a terrestrial laser data set of a
scene in a forest according to the individual tree branches.
REFERENCES
[Borgefors, 1996] Gunilla Borgefors, On digital distance trans-
forms in three dimensions, Computer Vision and Image Un-
derstanding, Vol.64 No.3 pp.368-376.
[Borgefors et al, 1999] Gunilla Borgefors, Ingela Nystróm and
Gabriella Sanniti Di Baja, Computing skeletons in three di-
mensions, Pattern Recognition 32 (1999) 1225-1236.
X
E 4
$ ^
^ "P
pee
+, Led *
;
ddr
Figure 8: Point cloud segmentation. Left: segmented skeleton;
right: detail
[Clark, 2002] N. E. Clark, 3D reconstruction of a tree stem us-
ing video images and pulse distances, Symposium on Statis-
tics and Information Technology in Forestry. September 8-12,
2002 Blacksburg, Virginia USA
[Lohou and Bertrand, 2001] Cristophe Lohou and Gilles
Bertrand, A new 3D 12-subiteration thinning algorithm based
on P-simple points, IWCIA, 2001.
[Palágyi, 2003] K. Paliágyi, J. Tschirren, M. Sonka: Quantita-
tive analysis of intrathoracic airway trees: methods and vali-
dation, LNCS 2732, Springer, 2003, 222-233.
[Palágyi et al, 2001] K. Palágyi, E. Sorantin, E. Balogh, A.
Kuba, C. Halmai, B. Erdóhelyi and K.Hausegger, A sequen-
tial 3D thinnig algorithm and its medical apllications, in M.
F. Insana and R. M. Leahy (Eds.): IPMI 2001, LNCS 2082,
Springer 2001, pp.409-415.
[Palágyi and Kuba, 1999] Kálmán Palágyi and Attila Kuba, Di-
rectional 3D thinning using 8 subiterations, LNCS 1568,
Springer, 1999, 325-336.
[Pfeifer and Gorte, 2004] Norbert Pfeifer and Ben Gorte, Au-
tomatic reconstruction of single trees from terrestrial laser,
IAPRS 35, Istanbul.
[Pyysalo and Hyyppae, 2002] Ulla Pyysalo, Petri Rônholm,
Hannu Hyyppae Henrik Haggré, Reconstructing tree crowns
from laser scanner data for feature extraction. IAPRS 34 Part
3-B pages B-218 ff (4 pages).
[Serra, 1985] J. Serra, Image Analysis and Mathematical Mor-
g S
phology, Academic Press, London, 1982.
[Sun, 2000] Guoging Sun, Modeling lidar returns from forest
canopies, TGARS Vol.38, No.6.
[Svensson and Borgefors, 2002] Stina Svensson and Gunilla
Borgefors, Distance transforms in 3D using four different
weights, Pattern Recognition Letters 23 (2002) 1407-1418.
[Telea and Vilanova, 2003] Alexandru Telea and Anna Vi-
lanova, À robust level-set algorithms for centerline extraction,
Joint EUROGRAPHICS-IEEE Symposium on Visualisation
(2003)
[Verwer, 1991] B.Verwer, Local Distances for Distance Trans-
formations in Two and Three Dimensions Pattern Recognition
Letters Vol.12, No. 11 (1991), 671-682
