Particle Tracking Velocimetry (PTV) was chosen as a
tool for the investigation of wave generated turbulences
and turbulent transport. The flow field was visualized by
small polystyrol particles in a light sheet illumination. The
particles are imaged by a CCD camera as streaks. By
tracking individual particle streaks from one frame to the
next the Lagrangian vector field (trajectories, particle path)
is extracted.
A 1-3cm thick light sheet parallel to the main wave prop-
agation direction is used to illuminate small (50-150 um in
diameter) polystyrol (LATEX) seeding particles. The depth
of the light sheet was chosen such that the particles stay in
the illuminated area long enough to enable tracking. The
light sheet is typically generated from below of the channel
by Halogen lamps. An area of typically 14.0 x 10.0 cm? is
imaged by a digital 200 Hz CCD camera (DALSA CA-DA-
0256) with a spatial resolution of 256 by 256 pixels. Due to
the movement of the particles during the exposure time of
zin ms, they are imaged as streaks. The image sequences
are stored on CD-Rom for later processing.
2.1 The Tracking Algorithm
Image Formation
Segmentation
Image Sequence
Analysis
Displacement
Vector Fields :
Figure 3: An overview over the different steps of image processing,
required for the evaluation of the vector field.
Several image processing steps are required for the extraction
of the flow field from the image sequences (see also Fig. 3).
First the particle streaks are segmented by a specially
developed segmentation technique. Each object is labeled
and finally the correspondence problem of identifying the
same object in the next image frame is solved by calculating
its streak overlap. Repeating this algorithm will track
segmented particles through the image sequence. Details
can be found in [Hering et al.,95a].
Segmentation The histogram (Fig. 4) of a a streak image
shows two distinct maxima, at the low gray values being faint
particle streaks and the background and at high gray values
being reflections at the water surface and bright particles.
Therefore the intensity of the streaks ranges from the very
low to the very high gray value. Simple pixel based segmen-
tation techniques cannot be chosen as the streak images do
not show a true bimodal distribution in the histogram. A re-
gion growing algorithm was developed for the discrimination
of individual particles from the background. Regions with
232
histogram
12000P
10000
8000f
6000
frequency
streaks
4000
2000 | |
Q ^ 71 4 ^ 209 ^ 35$
gray value
LS
7 LES
RITES
RL)
OLIS
30
Figure 4: Top: Original gray value image of polystyrol seeding
particles beneath the water surface. An area of 14.0 x 10.0 cm? is
imaged. Middle: Histogram of the above streak image. Although
appearing to show a bimodal distribution, particles cannot be seg-
mented by a threshold. Bottom: Pseudo 3d-plot of 32 x 32 pixels
of the original streak image. Streaks can clearly be identified as
local maxima in the gray value distribution.
similar features are to be identified and merged together to a
connected object.
Firstly the image g(x,y) is scanned through for local maxima
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
in the intens
mated by a |
search horizc
calculation o
gmin,i being 1
addition the
half width a
random nois
After these g
segments th
accepted as
higher than
gmin,i by int
the threshok
pixels formin
of the descri
Each object
with a flood
The size of e
large objects
Figure 5: Se
top). 501 obj
were eliminat
1:
ce
mm
Figure 6: 7
consecutive {
corresponding
