Istanbul 2004
|. In order to
re transformed
essed as ratio
1 plane (f+dc)
D:
(9)
distance was
m less than 30
| distance are
vith tube level.
imated to be +
vas carried out
or image size
distance are
the tree trunk.
ass. Error in À
y of image
cation plane,
geometrically
are compared
veen compass
|| axis) is 6? in
1racy compass
ind value was
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
4.4 Error in scale
[Images are in central projection and therefore objects in
different depths have different scales. Scale is correct only for
objects, which are in rectification plane like the scale bar. It is
assumed that tree crown is widest in rectification plane.
However, tree crowns are irregular in shape. Typical image
capture distance was 25 m and crown width 5 m. Therefore
distance from crown to the camera is varying from 22.5 to 27.5
m. In this interval the scale is changing from 0.006 to 0.0073. If
crown width measured from the image is 1200 pixels and its
real distance to the camera is 24 m instead the 25 m, error in
dimension is 36 cm.
S. DISCUSSION
5.1 Method advantages
In the study a photogrammetric method is suggested to measure
tree crown dimensions. The suggested method does not require
expensive instruments, such as terrestrial laser scanner or real-
time-kinematic GPS. Instruments required are digital camera,
camera platform, hypsometer, compass and reference scale bar.
Collecting reference data with the method is faster than with
traditional method. Capturing images does not require lifting
platform or climbing to the tree.
The automation level of image pre-processing is high because
relative orientation between image plane and rectification plane
is constant. Therefore, image rectification is batch -process.
Image rectification does not require known points in image area
or their signalisation.
5.2 Method restrictions
The terrestrial image based method has several limitations. The
basic problem is whether imaging from two directions is enough
for crown modelling? Measurements were not three-
dimensional. A three dimensional model could have been
extracted by traditional stereo imaging procedure. Reason for
not to do that is to keep imaging procedure simple and efficient.
Stereo measurements also require common points for relative
orientation and known points for outer orientation. With
deciduous trees, these points were difficult to find.
Images could be easily captured perpendicular to each other if
surrounding is open. Typically only few poses were available.
Also light direction excluded several image capture poses. It is
not realistic to assume that tree could be imaged in forest
surroundings from all directions.
5.3 Laser scanning accuracy
The method development initiated from the need to collect
reference material for laser scanned tree crown study. In laser
scanning, co-ordinates of the reflecting object are calculated
based on the known position of the scanner, pulse flight
direction and the distance between the scanning instrument and
the object. The main factors affecting accuracy of each point are
(a) range measurement accuracy, (b) instrument's positioning
accuracy and (c) laser beam direction accuracy (Baltsavias,
1999).
Accuracy of the range measurement is dependent on the
accuracy of pulse flying time measurement, the length of the
pulse and the shape of the echo (Katzenbeisser, 1998;
Baltsavias, 1999). The range measurement is most accurate
when whole pulse is reflected from one plain surface only,
which is in perpendicular position towards pulse flying
direction (Katzenbeisser, 1998). It is realistic to assume that in
tree crown laser pulse is not reflected from one surface only, but
from several small surfaces, such as leaves or needles. Pulse
echoes from different surfaces are separated only if their
distance is longer than pulse length. It is assumed that
planimetric accuracy of individual laser point is 20 cm.
Dimension measurement variance from one laser point to
another laser point is sum of single point measurement variance,
0,4 m.
6. CONCLUSIONS
In the study, a new method is suggested to measure tree crown
dimensions. The method’s accuracy was estimated based on the
rotation angle determination accuracy. Maximum errors of
vertical and horizontal measurements were 0.4 m, respectively.
However, measurement accuracy was usually better, because
tree is located in the middle of the image and largest errors take
place at image corners and edges (see Figure 7). The method is,
therefore, suitable for intended use, verification of laser scanned
tree crown evaluation.
Figure 7. Error surfaces and tree profile.
Accuracy of method may be improved with more accurate
rotation measurement. Level of tree crown shape reconstruction
may be increased if images are captured from more that two
directions.
ACKNOWLEDGEMENT
The author would like to thank Jenny and Antti Wihuri
Foundation, Foundation for Research of Natural Resources in
Finland and Finnish Academy for financial support.
REFERENCES
Ahokas, E., Kaartinen, H., Matikainen, L., Hyyppä, J., Hyyppâ,
H., 2002. Accuracy of high-pulse laser scanners for digital
target models. In: Observing our environment from space. New
solutions for a new millenium. Proceedings of the 21st EARSeL
Symposium, Paris, 14-16 May, 2001, pp. 175-178. Balkema
Publishers.
