See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226240853

Twenty years of particle image velocimetry

Article in Experiments in Fluids · August 2005

DOI: 10.1007/s00348-005-0991-7

CITATIONS READS

1,399 5,998

1 author:

Ronald J. Adrian
Arizona State University
392 PUBLICATIONS 37,418 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ronald J. Adrian on 03 June 2014.

The user has requested enhancement of the downloaded file.

Experiments in Fluids (2005) 39: 159–169
DOI 10.1007/s00348-005-0991-7

O R I GI N A L S

R. J. Adrian

Twenty years of particle image velocimetry

Received: 4 October 2004 / Revised: 25 March 2005 / Accepted: 25 March 2005 / Published online: 6 July 2005

Springer-Verlag 2005

Abstract The development of the method of particle special issues on PIV (Kompenhans and Tropea 1997;
image velocimetry (PIV) is traced by describing some of Adrian 2000, 2002) contain very useful collections of the
the milestones that have enabled new and/or better more recent work. Lastly, the reader can find many
measurements to be made. The current status of PIV is examples of state-of-the-art applications in this Special
summarized, and some goals for future advances are Issue.
addressed. The most rudimentary form of PIV could probably
be traced far back in history to the first time a person
possessing the concept of velocity watched small debris
moving on the surface of a flowing stream. For example,
Fig. 1 shows algae floating on the waters of a moat in
the backs of Trinity College, Cambridge, UK. It is
1 Historical development almost inconceivable that a great intellect like Isaac
Newton would not have observed the moving patterns
The year 2004 marked the 20th anniversary since the and seen the potential for visualizing and even measur-
term ‘‘particle image velocimetry’’ (PIV) first appeared ing the surface velocity from the displacements of the
in the literature. This article gives a personal view of the particles of algae. From this viewpoint, particle veloci-
development of PIV over those 20 years, followed by a metry is old and very simple. However, in its modern
summary of the current state-of-the-art and a prospec- form, PIV means the accurate, quantitative measurement
tive view of some of the improvements that are needed of fluid velocity vectors at a very large number of points
and the future possibilities in the field. The presentation simultaneously, and we now understand that this is,
reflects the author’s experiences and views of certain indeed, a very challenging, complicated, and relatively
developments that seemed particularly important or recent achievement.
interesting. No attempt has been made to make the The first investigators to achieve such measurements
presentation exhaustive, or to credit in any way that is actually used the method of laser speckle, originally
complete the many people who have advanced the field. developed in solid mechanics, and showed that it could
The reader can achieve a much more complete under- be applied to the measurement of fluid velocity fields. In
standing of the full scope of these developments by 1977, three different research groups, Barker and
referring to the excellent compilation of papers by Grant Fourney (1977), Dudderar and Simpkins (1977), and
(1994), to a very good book on PIV by Raffel et al. Grousson and Mallick (1977), independently demon-
(1998), and to the bibliography of PIV by Adrian (1996), strated the feasibility of applying the laser speckle phe-
which is almost exhaustive through 1995 and documents nomenon to fluid flow by measuring the parabolic
much prior work, including all of the first decade. Three profile in laminar tube flow. The principal elements of
their experiments were the use of double-exposure
photographs and planar laser light sheet illumination
R. J. Adrian and interrogation by forming Young’s interference
Laboratory for Turbulence and Complex Flow, fringes from the many pairs of displaced laser speckles in
Department of Theoretical and Applied Mechanics,
University of Illinois, Urbana, IL 61801, USA small interrogation spots on the specklegrams. By 1983,
a young doctoral student working at the v. Karman
Present address: R. J. Adrian
Department of Mechanical and Aerospace Engineering,
Institute, Belgium, Meynart (1979, 1980, 1982a, 1982b,
Arizona State University, Tempe, AZ 85287-6106, USA 1983a, 1983b, 1983c), was the leading practitioner of this
E-mail: [email protected] method, and he had shown that practical measurements
160

other using a dimensionless number called the source

density. The source density equals the mean number of
particles in a resolution volume, and the number of
overlapping images in the image plane can be expressed
in terms of it. For fluids, the allowable concentration of
scatterers is normally too small to produce source
densities large enough to have speckle patterns formed
by overlapping images. Higher particle concentrations
are either not achievable or not desirable fluid dynam-
ically (unless one intends to produce two-phase flow
effects). Hence, one almost always sees particle images
rather than speckles.
Many researchers became interested in PIV because it
offered a new and highly promising means of studying
the structure of turbulent flow. This goal strongly
influenced the choices made in the development of the
Fig. 1 Algae floating on the surface of water serve as flow markers
for elementary particle image velocimetry method. By its nature, turbulence is a phenomenon that
occurs over a wide range of physical scales, extending
from the largest scales of the flow down to the Kol-
could be made in laminar flow and turbulent flow of mogorov scale. Hence, a successful measurement tech-
liquids and gases, thereby stimulating intense interest nique must be able to measure over a wide dynamic
from the fluid mechanics community. range of scales in length and velocity. Another salient
While Meynart referred to his work as laser speckle feature of turbulence is its randomness, which may make
velocimetry (LSV), the images in his papers often con- it impossible to determine a priori the direction of flow.
tained images of individual particles instead of speckles. Hence, the measurement technique must be able to sense
The first explicit recognition of the importance of par- flows in all directions. Turbulence also occurs at high
ticle images was made in two short, contemporaneous Reynolds number, which often means high velocity.
papers by Pickering and Halliwell (1984) and Adrian Accelerations are large, and, therefore, the particles
(1984). In the latter, it was argued that the illumination must be small enough to follow the flow in the presence
of particles in fluid flows by a light sheet would seldom, of large local and randomly fluctuating accelerations.
if ever, create a speckle pattern in the image plane. This implies the use of very small particles, a few
Instead, the image plane would contain images of microns in size, and the small light scattering cross-
individual particles, such as those shown in Fig. 2. The section of such particles implies the use of high intensity
name particle image velocimetry (PIV) was proposed to illumination. Coupled with the short time exposures
distinguish this mode of operation from the laser needed to capture images of fine particles without
speckle mode. A simple criterion was defined by which blurring, these requirements lead naturally to the use of
one could predict the occurrence of one mode or the high intensity, pulsed lasers.

Fig. 2 Double-pulsed image

having low source density and
high image density. Flow is
from left to right
 161

Fig. 3 Typical monoscopic

particle image velocimetry
system

While these features were necessary for turbulent flow, velocimetry, there existed a good understanding of the
the capabilities they gave were also useful over a wide particle sizes needed to follow turbulent flows, and of
range of fluid flow problems. Consequently, the standard light scattering, so it was possible to compute, using Mie
basic PIV system now consists of a pulsed laser with a scattering theory, the exposure of images that would
light sheet illuminating particles a few microns in diam- result for appropriate particles. In particular, it was
eter in gases and, perhaps, a few tens of microns in liq- possible to show that pulsed lasers would provide
uids. The main option for recording the images is enough energy to obtain good photographic images
interline-transfer PIV video cameras, and interrogation from micron-sized particles in air and 10–30-lm-sized
by correlation analysis is a de facto standard, at least for particles in water. Subsequently, a big step in PIV
the present moment. A typical single-camera system is practice was to use double-pulsed solid-state lasers. They
shown in Fig. 3. These choices seem obvious now, but produced excellent double exposure photographs of
20 years ago, one was faced with choosing between particles without much limit on speed or fluid using
chopped continuous wave (CW) lasers, pulsed lasers, high-resolution (300 line/mm) film. The earliest use of
CW illumination with a shuttered recording camera, or Nd:Yag lasers appears to be in 1986 (Kompenhans and
xenon flash lamps. Then, there was also a variety of Reichmuth 1986). Still later, Nd:Yag lasers became
illumination coding sequences, including double-pulsed, available in compact, dual oscillator packages with self-
streak, streak and pulse, multiple-pulsed, and non-uni- contained cooling supplies, and they have become the
formly spaced pulses. In a review paper (Adrian 1986a), current workhorse of PIV.
the author once attempted to encompass systematically The idea of using auto-correlation of double-
all of the various possibilities for optical velocimetry by exposure images of multiple particles in small interro-
listing the leading candidates for various types of illu- gation spots, instead of measuring the spacing and
mination, coding, particles, image recording, and inter- orientation of Young’s fringes that form from illumi-
rogation. Given about three to five different candidates nating such spots, was first proposed in 1983 (Sutton
for each of these categories, there were several hundred et al. 1983; Adrian and Yao 1984).1 Details aside, the
combinations that might have produced potentially via- two approaches are related by Fourier transform,
ble systems. In the mid-1980s, the confusion engendered but the sharp signal peak in the correlation plane is
by this wealth of options was quite evident: one could
find dozens of papers describing different types of flow- 1
There may be many earlier papers from different fields that sug-
measuring systems that used optical imaging of particles, gested using correlation in somewhat different contexts. For
each differing from the other by their means illumination, example, Soo et al. (1959) presented a particularly prescient pro-
coding, particles, recording, and interrogation. posal for the ‘‘determination of turbulence characteristics of solid
particles in a two-phase stream by optical cross-correlation.’’ Leese
The energy necessary to illuminate fine particles and et al. (1971) describe ‘‘an automated technique for obtaining cloud
produce images of sufficient exposure and clarity was a motion from geosynchronous satellite data using cross-correla-
major issue in PIV. From experience with laser Doppler tion.’’
162

obviously the correct signal on which to base measure- Practically, it was impossible to perform two-dimen-
ments. In both methods, the entire image is divided into sional Fourier transforms or two-dimensional correla-
a grid of (usually overlapping) interrogation spots, and tion analysis on such machines. Therefore, there was
the particle images in each spot are interrogated to ob- considerable interest in non-statistical methods, such as
tain the mean displacement of the particles within each tracking particles individually. Alternatively, several
interrogation cell, which consists of the intersection of groups seriously pursued the determination of two-
the interrogation spot area, AI, and the thickness of the dimensional correlations by analog optical means
light sheet, Dz0. Analysis of the auto-correlation method (Morck et al. 1993; Vogt et al. 1996). Particle tracking
(Keane and Adrian 1992) led to the definition of a sec- implied operating with low image density so that the
ond dimensionless number, called the image density. It is probability of finding more than one pair of particles per
equal to the average number of scatterers in an inter- interrogation spot was small. Then, using the principle
rogation cell. This number proved to be very important that nearest-neighbor images corresponded to the same
in describing the characteristics PIV systems and in particle (which is only approximate for small, but finite
optimizing their design. The low image density limit image density), one could make successful measure-
corresponds to particle tracking, because, in that limit, it ments. The difficulty with this method was that, at the
is improbable to find more than one image pair per spot. reduced image density that accompanied reduced particle
The high image density limit corresponds to multiple concentration, the number of vectors per unit area was
particle correlation PIV (Fig. 4). not large enough to resolve turbulent fields completely.
In the first decade of PIV, the greatest challenge was To improve the spatial resolution, various investiga-
the interrogation of the images, simply because computer tors sought to optimize the low image density method by
capabilities were not adequate for the task. In 1985, the using interrogation windows of variable size, shape, and
DEC PDP 11/23 was a common digital computer in displacement. This led to the implementation of adap-
many fluids laboratories. It typically had 128 KB of tive windowing methods. Currently, adjustable window
RAM and a 30 MB hard drive. Imagine holding the methods enjoy use as a means of optimizing single-
operating system, the executable program, and the data exposed double-frame images obtained with digital
in a RAM space that is the same size as the minimum cameras.
document file size used by current word processors. At the time that Meynart performed his work using
Young’s fringes, the dynamic velocity range of the
technique, defined as the maximum velocity measurable
divided by the minimum velocity measurable, was
somewhere between 5 and 10. PIV was a velocity-mea-
suring instrument that had a 1-digit display! The
dynamic range was clearly far too small for the method
to be of value in serious fluid mechanics research. The
problem was that the dynamic range corresponds to the
maximum displacement of the images divided by
the minimum displacement that can be measured. In the
double-exposure images used at the time, the lower limit
was determined by the images overlapping when the
displacement was less than 1 image diameter. Thus, if
the maximum displacement was 10 image diameters, the
dynamic range was approximately 10:1.
The idea of applying an artificial spatial shift to the
second image was developed to improve the dynamic
velocity range and to provide a means of determining the
direction of the particle displacement from double-ex-
posed images (Adrian 1986b). In this method, the images
were recorded in such a way that the second image was
shifted precisely in a known direction so that the direc-
tion of flow could be determined unambiguously. Fur-
ther, the probability of two images from the same particle
overlapping was zero, and this solved the critical problem
of limited dynamic range. By eliminating the overlap of
particles images at small displacements, the dynamic
range immediately increased to somewhere between 100
and 200, where it remains to this day. Although
researchers continue to strive for a larger dynamic range,
it is now large enough to permit good measurements,
Fig. 4 Analysis of a grid of interrogation spots provided the PIV system is optimized.
 163

Various methods of interrogation by correlation, things happen: the mean number NI decreases, and the
including correlation of separately recorded exposures, fractions FI and F0 also decrease because fewer particles
have been investigated in theoretical/numerical simula- remain within the smaller volume for both exposures.
tion studies (Keane and Adrian 1992). The main issue is Decreasing the value of NIFIF0 by even a factor of two
whether or not the signal peak in the correlation plane from, say, 10 to 5 drastically reduces the fraction of valid
is larger or smaller than the random noise peaks. If it is vectors.
smaller, the wrong displacement is identified, and the One of the most important changes in the PIV tech-
measurement is invalid. The essence of the simulation nique was the move from photographic to videographic
studies was contained in a simple curve that collapsed recording. This change profoundly influenced the
the behavior of all the different systems—autocorrela- usability and, hence, the popularity of PIV. Of course,
tion, cross-correlation, variable window size, and mul- many researchers had been using digital cameras in
tiple-exposure systems—onto a single plot of the preference to film for years. For example, film recording
probability of a valid measurement versus a single was seld()om used in Japan. But, in the early 1990s,
dimensionless parameter (Fig. 5). This parameter char- several investigators, most notably Willert and Gharib
acterizes the effects of out-of-plane loss of images (F0 is (1991) and Westerweel (1993), published results indi-
the mean fraction of particles in an interrogation cell cating that the low resolution of digital cameras was not
that remain within the light sheet after displacing per- as serious an issue as others had supposed, and that
pendicular to the light sheet), in-plane loss of images (FI digital PIV could be accurate enough to provide useful
is the mean fraction of particles in an interrogation cell results. Photographic film possessed very high resolu-
that remain within the interrogation spot after displac- tion—100 line/mm for T-Max and 300 line/mm for
ing in the plane of the light sheet), measurement volume Technical Pan on 25·35 mm, or even 100·125 mm films.
size and shape, and particle concentration (as contained In comparison, digital camera resolution was typically
in NI). The non-dimensional parameter NIFIF0 is 500·500 pixels. However, digital cameras possessed high
essentially the mean number of particle image pairs per regularity in the location of the pixels relative to random
interrogation volume, taking into account the size of the locations of grains on a film, and clever methods were
volume due to windowing—a sort of generalized image developed to enhance the accuracy of the interrogation
density. The curve in Fig. 5 proves to be the single most of digital images. Moreover, the resolution of digital
important curve needed to optimize a PIV system, i.e., to cameras increased rapidly to 1,000·1,000 pixels, and
achieve a high probability of valid measurements. If the current 11-megapixel cameras are essentially equivalent
value of NIFIF0 is above 7–10, the probability of making to 100 line/mm 35 mm film.
valid measurements is very high—approaching 100%. In the early 1990s, it was clear that digital imaging
Then, straightforward interrogation by correlation would become the standard at some point in the future.
yields excellent results with a very high density of vec- What was perhaps not appreciated was the extent to
tors, like the field shown in Fig. 6. If, however, one were which digital imaging could simplify PIV and make it a
to reduce the size of the measurement volume, two process with which everybody was willing to work. The
work by Nishino et al. (1989) was extremely influential
in this regard. They presented the best turbulence sta-
tistics available from PIV at the time. They achieved
highly stable averages by taking over 19,200 video
images. This was far beyond anything one could do with
photographic film. The appeal of digital PIV rested not
only on the ease of acquiring images, but it also elimi-
nated the problem of mounting and carefully registering
each film frame on an interrogation table. The maximum
number of PIV photographs taken by even the most
determined investigators seldom, if ever, exceeded 1,000.
If one wanted good, accurate turbulence statistics, it was
necessary to use digital PIV. Hence, digital PIV enjoyed
increasing use in the mid-1990s, and now it is used
almost exclusively. The possibility of taking thousands
of PIV images made it desirable to speed up the inter-
rogation process and to automate the vector clean-up
process. Dantec developed and sold an impressively fast
hard-wired PIV correlator, but, ultimately, the incredi-
ble advance of PC capability and the flexibility of soft-
ware drove the development away from specialized,
Fig. 5 Probability of successful correlation analysis of an interro-
gation spot versus the image density, defined as the mean number
hard-wired devices.
of particles in the interrogation spot. Correlations between spots of The other outstanding impact of digital PIV came
different area are considered (Keane and Adrian 1992) with the advent of interline transfer cameras that could
164

Fig. 6 Vector field of flow

downstream of a rearward
facing step (expansion) in a pipe
flow. Vectors obtained by
gridded interrogation spot
autocorrelation analysis of a
100·120-mm double-exposed
photograph with image shifting
and at high image density
(Brouillette 1994)

hold two images recorded in rapid succession by trans- method have been proposed (see Proceedings of the
ferring the first image recorded by each pixel to an on- International Symposia on PIV 1999, 2001), all with the
chip storage well, and then record a second image. It is goal of extending the particle-tracking approach into the
the author’s understanding that the PIV community is realm of high image density. Figure 7 shows a typical
indebted to Lourenco et al. (1994) for convincing Kodak result of the super-resolution procedure by Takehara
to make such cameras for the PIV market. These cam- et al. (2001). The research groups of Yamamoto, Ko-
eras enabled three important improvements. First, it was bayashi, and Okamoto have each advanced the inter-
known theoretically that cross-correlation of separately rogation process considerably (see Okamoto et al. 1995;
recorded images of the first and second exposures was Song et al. 1999; Ishikawa et al. 2000; and the references
superior to the auto-correlation of double exposures cited therein). The approach of Yamamoto and co-
(Keane and Adrian 1992). But, cross-correlation could workers can be put in a class of interrogation methods
not be implemented conveniently until the new cameras that might be called ‘‘gridless correlation.’’ The idea is to
became available. Second, the cross-correlation cameras pick each particle and a surrounding group of 5–10
eliminated the need for image shifting: the direction of particles as a characteristic pattern. The particle plus
flow was determined automatically by the order of the group pattern is correlated from one exposure to the
exposures. Third, and most importantly, small dis- next. In principle, this could yield a vector for every
placement image overlap was eliminated completely, so particle, but of course, not all correlations yield a valid
that a large dynamic range was possible. The introduc- result. Even so, the method is very attractive. A third
tion of these cameras was one of the most important line of attack is the hierarchical correlation method, in
developments in the field of PIV. which correlation results from large interrogation spots
In PIV, ‘‘super-resolution’’ refers to means of inter- are used to guide correlation analysis of smaller spots,
rogation that improve the spatial resolution beyond that and so on, until very small spots are used (Hart 2000).
of the basic correlation interrogation spot. As first Yet another approach, based on correlation, is to rotate
proposed (Keane et al. 1995), the vectors from a stan- and strain the second window and to perform correla-
dard correlation analysis were used to enable reliable tion using six parameters: two translations, two rota-
image pairing in a particle-tracking scheme, thereby tions, and two strains (Huang et al. 1993). The approach
obtaining about 5–10 individual particle vectors for definitely yields a more accurate evaluation of the
each interrogation spot. Many improvements to this derivatives (see Lecordier and Trinite 2004).
 165

Fig. 7a, b Super-resolution

PIV. a Vector obtained on a
regular grid by correlation
analysis. b Vectors of individual
particles obtained by Kalman
filter particle tracking guided by
using the vectors in a as first
estimates (Takehara et al. 2001)

Many developments also occurred on the optical side not encounter this limitation because the projected vol-
of the PIV system. Stereographic imaging was used early ume of particles is much smaller. In this approach, one
to make photogrammetric measurements by particle can use ray tracing to determine the relationship between
tracking in volumes (Guezennec et al. 1994; Dracos et al. image plane locations and particle location (Arroyo and
1993; Maas et al. 1993; Kasagi and Nishino 1993). The Greated 1991) or generalized calibration with a target in
consensus experience is that the projection of particles the flow (Soloff et al. 1997). Stereographic PIV solves the
from 3D space onto 2D camera image planes creates problem of perspective error, as well as giving the third
particle image overlaps that limit the number of particles velocity component, and it has proven to be a practical
that can be imaged to about 3,000. Overlapping images generalization of monoscopic PIV.
could not be paired unambiguously. Recent work As noted above, the best performance achieved by 3D
(Pereira and Gharib 2002) using clever, out-of-focus photogrammetry yields about 104 vectors in a cubic vol-
imaging has pushed this number to about 104. Stereo- ume, corresponding to a little more than 20 vector sample
graphic imaging of particles in planar laser sheets does points per side. After accounting for the randomness of
166

the sample locations by dividing by p, the particle track-

ing velocimetry measurement is only equivalent to sam- 2 Current status
pling on a regular 7·7·7 grid. This may suffice for studies
involving relatively smooth flow fields or two-phase flow, Presently, the single-camera, planar light sheet, cross-
but it is not good enough for turbulence research. These correlation PIV with a double-pulsed Nd:Yag laser and
considerations have stimulated several efforts to make a 2,000·2,000-pixel cross-correlation PIV camera is the
volumetric PIV measurements from holographic record- standard system sold by commercial companies. Cool-
ings. Holographic recordings eliminate the particle image ing the cameras to achieve a higher signal-to-noise ratio
overlap problem because they make it possible to isolate for the images improves the effectiveness of each pixel,
one plane at a time. The consensus experience of various thereby, improving the effective resolution. In turbu-
research groups (Meng and Hussain 1991; Barnhart et al. lence research, just using simple 2D PIV has been
1994; Royer 1997; Trolinger et al. 1997) is that upwards of enormously rewarding in revealing fundamental aspects
106 regularly spaced vectors can be obtained using off- of the structure of turbulence. Some of these aspects
axis recording, with rather less using inline recording. had been inferred or guessed from earlier flow visuali-
This corresponds to a regular grid of about 100·100·100, zation, but the reliability of PIV visualization has made
which is as good as that commonly achieved in planar it possible to eliminate the guessing, to quantify vor-
PIV. The velocity accuracies are also comparable. Fig- ticity, and to reveal heretofore-unobservable phenom-
ure 8 shows a sample result. ena that allow completion of the structural pictures of
Why then, is holographic PIV not used more widely? certain canonical flows, such as wall turbulence. More
First, it is expensive; second, it requires considerable sophisticated forms of PIV will impact efforts to
skill; and third, one cannot realistically record enough understand turbulence, but one should not rush into
holograms to give stable turbulence statistics. This sit- complexity before mining the wealth of information
uation would change dramatically if electronically that can be had using 2D PIV.
readable and writable optical recording media were to Stereo PIV is now relatively common, and it is
become available with adequate resolution and sensi- working well, except that the out-of-plane component is
tivity. The current multi-mega-pixel cameras are already inherently less accurate than the in-plane components.
adequate for this task if one is willing to confine atten- Much of the focus over the last 5 years has been on
tion to a very small volume. Microscopic inline holog- developing accurate, robust means of measuring the
raphy has shown considerable promise (Jian et al. 2003). image displacement from the image field. It appears that
The adaptation of PIV to microscale flows (Santiago we are closing in on algorithms that are near optimum,
et al. 1998) reduced the typical PIV measurement vol- and that relatively little can be expected in terms of fu-
ume from 1 mm to 10 microns and less. This remarkable ture improvements in performance. The standard for 2D
two-orders-of-magnitude increase in spatial resolution is measurements is now about 300·300 vectors with a
achieved at the cost of reducing the field-of-view by a velocity dynamic range of no more than 200:1. Because
corresponding amount. Even so, it provides a useful new of this small dynamic range, many PIV experiments are
tool for microfluidics. still exercises in optimization. Framing rates have in-
creased dramatically with the introduction of new
cameras and high-repetition-rate lasers, and this devel-
Fig. 8 Three-dimensional vectors measured in a volume by off- opment offers a straightforward path for the expansion
axis, double-pulsed holographic PIV (Barnhart et al. 1994)
 167

of PIV capability. Coupling PIV with simultaneous • – Particle dynamics and the relationship between
planar laser-induced fluorescence (PLIF) has also en- measured particle displacement, particle velocity,
joyed success and seems relatively straightforward. and fluid velocity
Making combined measurements of fluid velocity and • – Imaging, including the accuracy and precision of
the velocity of a second phase such as particulate, mapping and distortion compensation
droplet or vapor phase is a viable and valuable extension • – Image recording and the effect of pixelization with
of PIV into multi-phase flow. The simultaneous mea- good noise models for the cameras
surements of liquid velocity and bubble phase by Lind- • – Optimum algorithms for locating particles with
ken and Merzkirch (2002) in Fig. 9 is an excellent maximum accuracy
example. • – Optimum algorithms for pairing particle images
with maximum reliability
• – Interpolating and smoothing regularly sampled
3 Desirable developments data from correlation interrogation or randomly
sampled data from particle tracking velocimetry or
It is risky to predict the future, especially when the super-resolution PIV
advance of PIV depends upon developments in the 2. New, more versatile particle seeding methods are
technology of components that lie outside the field, i.e., needed to:
computers, lasers, and cameras. However, one can, with
some confidence, list developments that would make • – Enable easy optimization of concentration and
PIV a more useful and incisive technique: higher concentrations in large volumes
• – Produce new particles for flows with severe accel-
1. A master theory should be developed that integrates eration—e.g., high-drag particles with large scat-
all of the following aspects of PIV: tering cross-sections, such as spiny spheres

Fig. 9 PIV measurements in a

flow containing five bubbles.
The bubble velocities are
indicated by the single arrows
(Lindken and Merzkirch 2002)
168

3. The goal should be set to achieve a velocity dynamic Adrian RJ (guest ed) (2000) Proceedings of the 3rd international
range of 1000:1—this would enormously increase the workshop on particle image velocimetry, Santa Barbara, Cali-
fornia, September 1999. Exp Fluids 29:S1–S307
utility of PIV and render tedious optimization of Adrian RJ (guest ed) (2002) Proceedings of the 4th international
experimental parameters less important symposium on particle image velocimetry, Göttingen, Ger-
4. The results of PIV experiments should be held to many, September 2001. Exp Fluids 33:735
increasingly rigorous standards. In particular, ex- Arroyo MP, Greated CA (1991) Stereoscopic particle image
velocimetry. Meas Sci Technol 2:1181–1186
perimentalists should routinely: Barker DB, Fourney ME (1977) Measuring fluid velocities with
• – Demonstrate the adequacy of the spatial resolu- speckle patterns. Opt Lett 1:135–137
Barnhart DH, Adrian RJ, Papen GC (1994) Phase conjugate
tion by performing grid resolution tests and/or holographic system for high resolution particle image veloci-
spatial frequency response tests metry. Appl Optics 33:7159–7170
• – Demonstrate the accuracy and reproducibility of Brouillette JN (1994) Incompressible turbulent flow through a
the velocity measurements sudden axisymmetric expansion using particle image velocime-
try. MS thesis, Department of Mechanical Engineering, Uni-
• – Routinely examine the probability density histo- versity of Illinois, Urbana-Champaign, Illinois
grams of the velocity data for evidence of experi- Dracos T, Virant M, Maas HG (1993) Three-dimensional particle
mental artifacts tracking velocimetry based on photogrammetric determination
of particle coordinates. In: Cha SS, Trollinger JD (eds) Optical
5. Means should be sought to reduce total system costs diagnostics in fluid and thermal flow. SPIE, Bellingham,
by: Washington, pp 457–466
Dudderar TD, Simpkins PG (1977) Laser speckle photography in a
• – Reducing the costs of light sources and cameras fluid medium. Nature 270:45–47
• – Developing low-cost, restricted-purpose systems, Grant I (ed) (1994) Selected papers on particle image velocimetry.
such as probe-PIV Milestone Series MS99, SPIE international society of optics
engineering, Bellingham, Washington
The reader will undoubtedly have some favorite Grousson R, Mallick S (1977) Study of flow pattern in a fluid by
scattered laser light. Appl Opt 16:2334–2336
developments to add to this list. Guezennec YG, Brodkey RS, Trigui J, Kent JC (1994) Algorithms
for fully automated three-dimensional particle tracking veloci-
Acknowledgements This paper was prepared with the support of metry. Exp Fluids 17:209–219
the U.S. Department of Energy, Los Alamos National Laboratory. Hart DP (2000) PIV error correction. Exp Fluids 29:13–22
The author would like to take this opportunity to acknowledge Huang HT, Fiedler HE, Wang JJ (1993) Limitation and improve-
with gratitude the many contributions of the students, post-doc- ment of PIV. Part I. Limitation of conventional techniques due
toral associates, senior visiting scholars, and academic collabora- to deformation of particle image patterns. Exp Fluids 15:168–
tors who, over the past 20 years, have studied particle image 174
velocimetry in the Department of Theoretical and Applied Me- Ishikawa M, Murai Y, Wada A, Iguchi M, Okamoto K,
chanic’s Laboratory for Turbulence and Complex Flow. They are, Yamamoto F (2000) A novel algorithm for particle tracking
in rough chronological order: C.-S. Yao, C. Landreth, J. Kom- velocimetry using the velocity gradient tensor. Exp Fluids
penhans, M. Lee, R. Keane, Z. C. Liu, D. L. Reuss, T. J. Hanratty, 29:519–531
P. Offutt, A. Prasad, M. Cui, C. Wark, H. Xu, T. Urushihara, J. Jian S, Malkiel E, Katz J (2003) Single beam two-views holo-
Westerweel, C. Westergaard, C. Meinhart, K. Paschal, J. Brouil- graphic particle image velocimetry. Appl Optics 42:235–250
lette, Y. Zhang, D. Barnhart, G. Papen, V. Troy, T. Oakley, E. Kasagi N, Nishino K (1993) Flow field measurement with particle
Loth, C. Tomkins, S. Soloff, U. Ullum, K. Christensen, R. Fer- image velocimetry. JSME 96:34–39
nandes, K. Sharp, N. Fujisawa, C. Kaehler, K. Nishino, J. Sak- Keane RD, Adrian RJ (1992) Theory of cross-correlation analysis
akibara, K.C. Kim, D. Hill, K. Takehara, J. Santiago, M. Olsen, Z. of PIV images. Appl Sci Res 49:191–215
Deng, B. Balasubramanian, E.Yamaguchi, M. Murphy, W. Lai, G. Keane RD, Adrian RJ, Zhong Y (1995) Super-resolution particle-
Elliott, and P. Vedula. imaging velocimetry. Meas Sci Technol 6:754–768
Kompenhans J, Reichmuth J (1986) Particle imaging velocimetry in
a low turbulent wind tunnel and other flow facilities. In: Pro-
ceedings of the IEEE Montech’86 conference, Montreal, Can-
References ada, October 1986. AGARD conference proceedings no. 399,
paper 35
Adrian RJ (1984) Scattering particle characteristics and their effect Kompenhans J, Tropea C (guest eds) (1997) Special issue: particle
on pulsed laser measurements of fluid flow: speckle velocimetry image velocimetry. Meas Sci Technol 8:1379–1583
vs. particle image velocimetry. Appl Opt 23:1690–1691 Lecordier B, Trinite M (2004) Advanced PIV algorithms with
Adrian RJ (1986a) Multi-point optical measurement of simulta- image distortion—validation and comparison from synthetic
neous vectors in unsteady flow—a review. Int J Heat Fluid images of turbulent flow. In: Stanislas M, Westerweel J,
Flow 5:127–146 Komenhans J (eds) Particle image velocimetry: recent
Adrian RJ (1986b) An image shifting technique to resolve direc- improvements, proceedings of the EUROPIV 2 workshop,
tional ambiguity in double-pulsed laser velocimetry. Appl Zaragoza, Spain, March/April 2003. Springer, Berlin Heidel-
Optics 23:3855–3858 berg New York
Adrian RJ (1996) Bibliography of particle velocimetry using Leese JA, Novak CS, Clark BB (1971) An automated technique for
imaging methods: 1917–1995. TSI Inc., St. Paul, Minnesota. obtaining cloud motion from geosynchronous satellite data
Also available as TAM report no. 817, UILU-ENG-96-6004, using cross-correlation. J Appl Meteorol 10:118–132
Department of Theoretical and Applied Mechanics, University Lindken R, Merzkirch W (2002) A novel PIV technique for mea-
of Illinois, Urbana-Champaign, Illinois surements in multiphase flows and its application to two-phase
Adrian RJ, Yao CS (1984) Development of pulsed laser veloci- bubbly flows. Exp Fluids 33:814–825
metry (PLV) for measurement of fluid flow. In: Patterson G, Lourenco LM, Gogineni SP, Lasalle RT (1994) On-line particle
Zakin J (ed) Proceedings of the 8th biennial symposium on image velocimeter: an integrated approach. Appl Opt 33:2465–
turbulence, Rolla, Missouri, September 1984, pp 170–186 2470
 169

Maas HG, Gruen A, Papantoniou D (1993) Particle tracking Pickering CJD, Halliwell N (1984) LSP and PIV: photographic film
velocimetry in three-dimensional flows. Part 1. Photogram- noise. Appl Opt 23:2961–2969
metric determination of particle coordinates. Exp Fluids Raffel M, Willert C, Kompenhans J (1998) Particle image veloci-
15:133–146 metry, a practical guide. Springer, Berlin Heidelberg New York
Meng H, Hussain F (1991) Holographic particle velocimetry: a 3D Royer H (1997) Holography and particle image velocimetry. Meas
measurement technique for vortex interactions, coherent Sci Technol 8:1562–1573
structures and turbulence. Fluid Dyn Res 8:33–52 Santiago JC, Meinhart CD, Wereley ST, Beebe DJ, Adrian RJ
Meynart R (1979) Flow velocity measurement by a speckle method. (1998) A particle image velocimetry system for microfluidics.
In: Proceedings of the 2nd European congress on optics applied Exp Fluids 25:316–319
to metrology, Strasbourg, France, November 1979, vol 210, pp Soloff S, Adrian RJ, Liu ZC (1997) Distortion compensation for
25–28 generalized stereoscopic particle image velocimetry. Meas Sci
Meynart R (1980) Equal velocity fringes in a Rayleigh–Benard flow Technol 8:1441–1454
by a speckle method. Appl Opt 19:1385–1386 Song X, Yamamoto F, Iguchi M, Murai Y (1999) New tracking
Meynart R (1982a) Convective flow field measurement by speckle algorithm of PIV and removal of spurious vectors using Dela-
velocimetry. Rev Phys Appl 17:301–305 unay tessellation. Exp Fluids 26:371–380
Meynart R (1982b) Digital image processing for speckle flow Soo SL, Tien CL, Kadambi V (1959) Determination of turbulence
velocimetry. Res Sci Instrum 53:110–111 characteristics of solid particles in a two-phase stream by
Meynart R (1983a) Instantaneous velocity field measurement in optical autocorrelation. Rev Sci Instrum 30:821–824
unsteady gas flow by speckle velocimetry. Appl Opt 22:535–540 Sutton MM, Wolters WJ, Peters WH, Ranson WH, Macneill SR
Meynart R (1983b) Speckle velocimetry study of vortex pairing in a (1983) Determination of displacements using an improved
low-Re unexcited jet. Phys Fluids 26:2074–2079 digital correlation method. Opt Eng 1:113–139
Meynart R (1983c) Speckle velocimetry: an application of image Takehara K, Adrian RJ, Etoh GT, Christensen KT (2001)
analysis techniques to the measurement of instantaneous A Kalman tracker for super-resolution PIV. Exp Fluids 29:S34–
velocity fields in unsteady flow. In: Proceedings of the inter- S41
national congress on instrumentation in aerospace simulation Trolinger JD, Rottenkolber M, Elandaloussi F (1997) Develop-
facilities (ICIASF’83), Haut-Rhin, France, September 1983. ment and application of particle image velocimetry techniques
IEEE record, pp 30–36 for microgravity applications. Meas Sci Technol 8:1573–
Morck T, Andersen PE, Westergaard CH (1993) Processing speed 1580
of photorefractive optical correlators in PIV-processing. In: Vogt A, Reichel F, Kompenhans J (1996) A compact and simple
Adrian RJ, Durao DFG, Durst F, Heitor MV (eds) Laser all optical evaluation method for PIV recordings. In: Adrian
techniques and applications in fluid mechanics. Springer, Berlin RJ, Durao DFG, Durst F, Heitor MV, Maeda M, Whitelaw
Heidelberg New York, pp 227–242 JH (eds) Developments in laser techniques and applications to
Nishino K, Kasagi K, Hirata M (1989) Three-dimensional particle fluid mechanics. Springer, Berlin Heidelberg New York, pp
tracking velocimetry based on automated digital image pro- 423–437
cessing. ASME J 111:384–391 Westerweel J (1993) Analysis of PIV interrogation with low pixel
Okamoto K, Hassan YA, Schmidl WD (1995) New tracking algo- resolution. In: Cha SS, Trolinger JD (eds) Optical diagnostics in
rithm for particle image velocimetry. Exp Fluids 19:342–347 fluid and thermal flow. SPIE, Bellingham, Washington, pp 624–
Pereira F, Gharib M (2002) Defocusing digital particle image 635
velocimetry and the three-dimensional characterization of Willert CE, Gharib M (1991) Digital particle image velocimetry.
two-phase flows. Meas Sci Technol 13:683–694 Exp Fluids 10:181–193

View publication stats