A 3D-SENSOR FOR THE MEASUREMENT OF PARTICLE CONCENTRATION FROM IMAGE SEQUENCES
Peter Geißler *
* Interdisciplinary Center for Scientific Computing
University of Heidelberg
INF 368, D-69120 Heidelberg, FRG
Peter.Geissler@iwr.uni-heidelberg.de
Bernd Jähne *?
? Scripps Institution of Oceanography
Physical Oceanography Research Division
La Jolla, CA 92093-0230, USA
bjaehne@giotto.iwr.uni-heidelberg.de
Commision V, Working Group V/III
KEY WORDS: Camera,Depth-from-Focus, Experiment, Optical,Sensor, Three-dimensional, Underwater
ABSTRACT
A sensor for the measurement of the concentration and size distribution of small particles is described. The particles are
visualized by a light blocking technique and imaged by a CCD camera. The concentration measurement is based on a depth-
from-focus approach to determine the positions of the particles in three-dimensional world coordinates. From these coordinate
sets of all observed particles both the measuring volume as well as the concentration can be determined. The sensor has been
developed for the underwater measurement of size distributions of air bubbles submerged by breaking waves to investigate the
influence of air entrainment to air-sea gas exchange.
1 INTRODUCTION
The investigation of the exchange of clima-relevate trace
gases such as CO» and oxygen is a current research topic
in modern oceanography. The underlying physical processes
are still poorly understood. At low windspeeds, gas exchange
is controlled by molecular diffusion through the air/sea phase
boundary. With increasing windspeed the situation changes
dramatically. Breaking waves entrain air in form of gas bub-
bles directly to the water. During their lifetime underneath
the water surface, these bubbles form an additional reservoir
which exchanges its gas contents with the water. Indeed, a
strong increase of the gas exchange rates is found when waves
start to break [Merlivat,83]. The influence of the air bubbles
strongly depends on parameters such as life time, penetration
depth and the amount of gas confined. All these parameters
are determined by the size of the bubbles. Therefore precise
measurements of bubble size spectra and their dependence
on wind speed are necessary.
To investigate the properties of air bubble distributions, we
have developed an optical sensor which images bubbles on a
CCD camera. This allows for a instrument which causes only
low flow disturbances. The need to place the measurement
volume as far away from the instrument as possible and to
image bubbles in a size range of 20 um to 1000 um with a
suitable resolution causes a narrow depth of field. Therefore,
blur in the images is introduced in a natural way, causing only
bubbles near the focal plane to be imaged sharply. With in-
creasing distance from the focal plane (subsequently denoted
by ’depth’) the images of the bubbles become blurred. With
our technique this blur is used to determine their depth and,
consequently, their three-dimensional position in real world
coordinates. From these data, both the measuring volume as
well as the size distribution is determined.
This paper will describe the sensor and the image processing
techniques used to determine the size distributions from the
image sequences. The sensor is also suitable for other types
of particles in gases or liquids, provided they are of uniform
shape and their refractive index is different from that of the
surrounding medium.
194
2 DEPTH-FROM-FOCUS BASED
CONCENTRATION MEASUREMENT
2.1 Introduction to Depth-from-Focus
The processes of imaging a three-dimensional scene onto a
two-dimensional sensor at a first glance causes the loss of
depth information. Obviously depth recovery needs additional
information. For example, stereo imaging can be used to
partly reconstruct the distance of objects from two images.
Another approach is the use of blur in the images for depth
reconstruction. In contrast to a pinhole camera which images
every point in object space to an ideal point on the image
plane, a lens with finite aperture images only object points
on the focal plane to points on the image plane. Increasing
distance from this plane then causes an increase of blur in
the image (Fig. 1).
Depth-from-focus techniques estimate the grade of blur and
correlate it with the distance of the objects from the focal
plane. The depth resolution depends on the change of blur
with the distance from the focal plane and is therefore related
to the depth of field of the optics. Thus, a small depth of
field results in a good depth resolution of a depth-from-focus
system.
image plane lens
focal plane
Figure 1: Defocused imaging: an object point at ug is imaged
to a point at the conjugated image plane at vo. A point at u # uo
results in a brightness distribution of width o on the image plane.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
Defocused im:
the well focus
with the point
of a point obj
Thus
spatial dor
Fourier doi
with O(Z) be
ing image. In
with the Four
function. Be
tion, it is not
the object and
The same ame
edges and a v
PSF. But mea
width of the F
of apparent b
from-focus ap
by several im:
problem. As a
quences ([Ens
tioned here. T
ferent parame
of the PSF ir
setup would c
finite aperture
contain blur c;
the object prc
reconstructed
systems with :
For the invest
results in sigr
used. For si
which need or
is only possibl
or the objects
other technica
ori known. T
uses the knov
pute an objec
2.2 Concen
The calculati
of particles re
observed part
ing volume. L
means of mec
to handle vol
cannot be ney
following app
The basic ide:
surements is t
era looks free
by its depth of
Both particle:
cles located f
then results ir
positions of 2
position perp
the gray valu
