74 
A ROBUST TECHNIQUE FOR TRACKING PARTICLES OVER LONG 
IMAGE SEQUENCES 
Frank Hering, Michael Merle, Dietmar Wierzimok! 
Institute for Environmental Physics, University of Heidelberg 
Im Neuenheimer Feld 366,D-69120 Heidelberg, Germany 
Phone: (49) 06221-563403, email: fhering@dali.uphys.uni-heidelberg.de 
! Now affiliated to IBM Research Center Heidelberg, FRG. 
Bernd Jähne 
Interdisciplinary Center for Scientific Computing 
University of Heidelberg, Im Neuenheimer Feld 368, D-69120 Heidelberg, Germany 
and 
Scripps Institution of Oceanography 
Physical Oceanography Research Division, University of California 
La Jolla, CA 92093-0230, USA 
KEY WORDS: Particle Tracking Velocimetry, Flow Visualization, Correspondence Problem 
ABSTRACT: 
À particle tracking technique at high particle concentration for the evaluation of flow fields beneath water 
waves is described. A 1-4 cm thick light sheet parallel to the main wave propagation direction was used to 
illuminate small polystyrol seeding particles. The depth of the light sheet was chosen such that the particles 
stay long enough in the illuminated area to enable tracking. An area of up to 14.0 x 9.0 cm? is observed by 
a digital CCD camera. The recording of continuous image sequences at up to 60 fields/s allow an extensive 
study of the flow field. An automatic tracking algorithm is described allowing particles to be followed over 
more than 200 images at particle concentrations up to 800 particles/image (image resolution 480 x 512 
pixel), yielding both Lagrangian and Eulerian vector field. The error of the computation of a velocity vector 
is usually below 3 96. 
1 INTRODUCTION 
Most particle tracking techniques use streak photography [7] as a tool for the determination of the flow field. 
The velocity field can be obtained by measuring length, orientation and location of each streak [4]. The length 
is commonly calculated from the end points of a streak, detected by a segmentation algorithm. This approach 
to particle tracking is only feasible at low particle concentration up to typically 30-100 particles/image [4] 
[1]. Most authors use a physical model as an interpolation scheme for identifying the same particle in the 
next image frame [1] [3]. This paper describes a novel particle tracking algorithm, allowing to track up to 800 
particles/frame using the image field overlap as the primary feature for solving the correspondence problem. 
2 OPTICAL SET-UP 
A 1-4cm thick light sheet parallel to the main wave propagation direction is used to illuminate small LATEX 
seeding particles. The light sheet is typically generated from beneath of the channel by Halogen lamps (see 
Fig. 1). The visualization used in Delft and in Heidelberg are basically identical and have been described 
earlier by [12] and [11]. An area of typically 14.0 x 9.0 cm? is imaged by a 60 Hz CCD camera (Pulnix TM-740 
and Sony XC75) with a spatial resolution Due to the movement of the particles during the exposure time, 
they are imaged as streaks (Fig. 2). The image sequences are stored on a Laser Video Disc (Sony LVR 5000) 
for later evaluation. 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop “From Pixels to Sequences’, Zurich, March 22-24 1995