202 
A HIGH PERFORMANCE SYSTEMFOR 3-DIMENSIONALPARTICLE TRACKING VELOCIMETRY IN 
TURBULENT FLOW RESEARCH USING IMAGE SEQUENCES 
T.Netzsch, BASF AG, ZI/CI, Q920. D-67056 Ludwigshafen, Germany. 
Phone: +49-621-60-72592, Fax: +49-621-60-71417, Email: pdsmgr@zi.basf-ag.de 
B.Jahne, IWR, University of Heidelberg, D-69120 Heidelberg 
Phone: +49-6221-562827, Fax: +49-6221-563405, Email: bjaehne@davinci.iwr.uni-heidelberg.de 
KEY WORDS: 3D-PTV, Stereo, flow measurement 
ABSTRACT: 
The standardmethodin turbulent flow measurement is Laser Doppler Anemometry, which gains only 2-dimen- 
sionalinformation about the velocity with good accuracy, but only for stationary points. 3D-PTV is able to cover 
all three components of a spatial flow field using stereoscopic image sequences. The determination of the 3rd 
component of the velocity vector requires the solving of the stereo correspondence problem. 
. Incontrast to other methods which solve first the stereo correspondence for single images and do the particle 
tracking as a second step, this paper describes a high-performance algorithm working on image sequences, 
where the first step is the particle tracking and then in a second step the stereo correspondence problem is re- 
solved for particle trajectories not for single particles. The advantage of this approach are the ability to track 
the particles for long time intervalls (7-10 s and more) while the stereo correspondence has a matching rate 
between 90% and 100% at particle densities up to750 particles and still a matching rate of 65% at 2000 partic- 
les. The typical computation times on a Silicon Graphics Indy (MIPS R4000, 100MHz) are 0.25 seconds/image 
for the particle tracking and 10 seconds for the stereo correspondence with sequences of 150 images and 350 
particels. These are considerable short times. 
The system has been developed using real image sequences. In a second step simulated data has been evaluated 
andit can be shown, that the limitations of the system come from the particle tracking. Thus further study will 
focus on improvement of the particle tracking. iC 
1 INTRODUCTION 
Three dimensional flow measurements have a wide area of interest. Theu are used in industrial applications as 
well as in basic research. Of special interest is the study of turbulence an transport processes [Já-85], where 
image processing methods are able to gain the needed information. While Particle Imaging Velocimetry allows 
the determination of 2-dimensional flow fields, 3-dimensional PTV is able to cover all three components of a 
spatial flow field using stereoscopic image sequences. Also PIV is not able to observe trajectories, which can be 
measured with 3-D PTV for longer sequences with an evident smaller amount of calcaulating operations. 
To obtain spatial information, the stereo correspondence problem has to be solved. Working on single images 
(Adamczyk 1988 and Maas1992) gives ambiguities, getting more severe with increasing particle density. 
Thus the stereo correspondence problem is solved comparing image sequences, whereby the stereo corre- 
spondence is achvied for particle traces in image sequences instead of single particles in single images. This 
new approach has the advantage of reducing the ambiguities and thus allows to work with high particle densi- 
fies(> 2000 Particles) while at the same time the amount of calculations is minimized and therefore the algo- 
rithm is considerable fast. 
The system has been developed using real image sequences. Simulated data has been computed for optimiza- 
tion and to find out the limitations of the system 
2 PRINCIPLE OF THE SYSTEM 
The flow is visualized using polystyrol particles ranging from 50 to 150 pum in diameter. After illuminating an ob- 
servation volume image sequences are recorded using two cameras. The image analysis includes: : 
- calibration of the camera system 
- 2D-particle tracking 
- solving the stereo correspondence problem for particle traces 
- determination of spatial velocity data 
This paper focuses on the stereo correspondence algorithm and the simulation. The calibration technique and 
the particle tracking are described in (Netzsch 1992, Posch 1990, Tsai 1986, Hering 1995, Wierzimok 1992). 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences", Zurich, March 22-24 1995