International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
In Haala (2003) terrestrial laser scanning was combined with
panoramic images and surface of tree trunks were extracted to
analyse tree species. Digital camera has been used to measure
individual tree variables from the digital images in the field
(Juujárvi et al. 1998). Lee et al. (2003) presented automated
methods of tree boundary extraction and Clark et al. (2000)
digital terrestrial photogrammetric methods for tree stem
analysis.
The objective of this paper was to develop a photogrammetric
method based on terrestrial digital images in order to measure
the geometric shape and derive parameters of individual trees.
The method does not require special instruments, time
consuming tacheometer measurement or positioning of image
capture location. Equipments required are a digital camera,
camera platform, hypsometer, compass, reference scale bar and
measurement tape. Error sources of suggested method are
analysed separately. This digital-image-based method will be
used to verify previously developed geometric 3D tree model
(Pyysalo & Hyyppà 2002).
2. MATERIALS
2.1 Test area, field measurements and laser acquisitions
The test area locates in urban environment in the city of Espoo,
Finland. Area is park type university campus with lots of
vegetation. 30 trees (pines, spruces and birches) were chosen as
test trees. 16 trees were imaged from one direction only and 14
trees from two directions. Trees were chosen from group of 50
trees. Tacheometer measurements have been carried out to
locate tree trunk positions and digital elevation model profiles.
Area has been scanned with airborne laser twice. The first one
was measured from aeroplane with German Toposys instrument
in late spring 2000. Flying height was 400 m, scanning
frequency 80 0000 Hz and cone diameter of laser footprint 20
cm in ground. Swedish Topeye instrument measured same area
in late summer 2002 from helicopter flying in 200 meters. Pulse
cone diameter in the ground was 20 cm.
2.2 Instrument
Images were taken using Nikon E-10 digital camera. Camera
was mounted to camera platform with flexible elevation and
rotation possibilities. Maximum image size, 3008x2000, pixels
was utilized and fixed size of camera aperture. During the
measurement camera was manually focused to the eternity,
which is also the same position the camera was calibrated.
3. METHOD
3.1 Terrestrial imaging
Several tests were carried out in the field to find optimal image
capture workflow. The crown dimension measurement task
required that both the base and tree crown top would be visible
in the same image. In the same time image should be captured
as close as possible for detailed measurement and also because
during the tests it was observed that even in sparse forests other
trees or other objects were typically shadowing the target tree if
the image shooting distance was too long. The camera stand
was approximately 1.5 meters high and imaged trees were 20-35
meters in height.
The applied image capture workflow was as follows:
e The image capturing location was chosen from 20-30
meter distance from the tree.
e Camera stand was levelled according to bubble level.
* The optical axis of camera was levelled horizontally
(® = 0).
e Angle K was set to zero using tube level.
e Reference scale bar was attached to the tree trunk.
e Camera was rotated around y-axis (¢) in the position
where trunk was in the middle of the image.
e (Q was set to 20° angle was measured using
hypsometer.
e After capturing the image angle ¢ was measured with
compass.
e Distance to the tree was measured.
Figure 1. Image capture. Camera was rotated in the position
where trunk was in the middle of the image (up left). Reference
scale bar was attached to tree trunk (up right). Camera was
rotated 20? (bottom left). Distance to the tree was measured
(bottom right.)
After imaging a tree the same process was repeated from
another location for the same tree so that image capture
directions were perpendicularly towards each other.
3.2 The image pre-processing
In the image pre-processing distortion errors were removed
utilizing camera calibration file (Figure 3). The calibration was
carried out for shortest focal length f, which was utilized in
image capturing. The images where rectified to plane parallel to
the trunk (Figure 2).
Interna
Figur
Based :
the rect
the im:
image |
square
of proj
et al, 2
where
After re
rectifica
and rec
transfor
rectifica
Figure