af tree. The
voxel space,
es it should —
] points of the
- 90 cylinders
.s.e. of 1.1cm
racy (r.m.s.e.)
niferous trees
; are shown (5
'en a different
and the hulls
was selected
on was not ac-
n higher parts
er running the
matically and
ts of cylinder
; shown in the
he slices were
(22.5° each).
tree axes.
rell, but it can
left tree was
ided by lower
der following
f the cylinder
"ss automatic.
der following
:sults in short
ing the work.
Istanbul 2004
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
, Vol XXXV, Part B5. Istanbul 2004
Figure 6: Point cloud from three coniferous trees and automatically reconstructed stems and outer hulls.
5 CONCLUSION
We presented a set of algorithms for the reconstruction of in-
dividual trees from dense laser scanning point clouds in forests.
Terrestrial laser scanning is, as this example shows, not only com-
peting with image based methods, but allows also to open new
application fields. The technology of laser scanning and the im-
mediate processing steps (i.c., registration) are not completely fit
for difficult and rough environments as forests, yet.
The cylinder models of the individual tree stems are being used
by forest managers and timber industry for determining important
parameters describing timber quality like taper, sweep and lean
which are closely related to the proportion of reaction wood. The
reconstructed crowns or outer hulls of the crowns, respectively,
allows the calculation of the crown projection area which one the
one hand describes the vitality or the competitive status of forest
trees and on the other hand is an important criteria for target-
oriented management of forest stands (Biging and Dobbertin,
1995),
It has been demonstrated that fully automatic modelling is possi-
ble, but visual interpretation of the point cloud allows to extract
more information, than the models currently deliver. As was to be
expected, reconstruction works better for lower, thicker branches,
where the point coverage is denser. The different threshold val-
ues should be set automatically from the data. Future research
direction aims therefore at i) extracting more branches, and ii) do-
ing away with the restriction of circular cross-sections. This will
allow the extraction of more information (c.g. quality measures
for wood production) from the laser scanner point cloud.
REFERENCES
ANN, 2003. ANN, approximate nearest neighbors.
http://www.cs.umd.edu/~mount/ANN/ (accessed 1
May 2004).
Besl, P. and McKay, N., 1992. A method for registration of
3D shapes. IEEE Transactions on Pattern Analysis and
Machine Intelligence, 14(2):239 — 256.
Biging, G. and Dobbertin, M., 1995. Evaluation of competition
indices in individual tree growth models. Forest Science,
41(2):360-377.
Frohlich, C., Mettenleiter, M., Hirt, F., Dalton, G., and Hines,
D., 2000. Imaging laser radar for 3-D modelling of real
world environments. Sensor Review, 20(4):273-281.
Gorte, B. and Pfeifer, N., 2004. Structuring laser-scanned
trees using 3d mathematical morphology. International
Archives of Photogrammetry and Remote Sensing,
Vol. XXXV, Istanbul, Turkey.
Lancaster, P. and Salkauskas, K., 1986. Curve and surface
fitting, An Introduction. Academic Press.
Lukacs, G., Martin, R., and Marshall, D., 1998. Faithful least-
squares fitting of spheres, cylinders and tori for reliable
segmentation. In Computer Vision. Proceedings of the 5th
European Conference on Computer Vision, pages 671—
686, Feiburg, Germany.
Pyysalo, U. and Hyyppä, H., 2002. Reconstructing tree crowns
from laser scanner data for feature extraction. Interna-
tional Archives of Photogrammetry and Remote Sensing,
Vol. XXXIV, 3A, Graz, Austria.
Shakarji, C. M., 1998. Least-squares fitting algorithms of
the NIST algorithm testing system. Journal of Research
of the National Institute of Standards and Technology,
103(6):633—641.
