Geoscience Reference
In-Depth Information
regimes. In the context of atmospheric and oceanographic
studies, similar 3D PTV systems have been recently imple-
mented to investigate, for instance thermal convective
flows [ Ni et al. , 2012] as well as the influence of Corio-
lis force on the transport properties of rotating flows [ Del
Castello and Clercx , 2011]. We detail in the following how
such multicamera tracking systems work.
15.2.1.1. Principle. The principle of optical particle
tracking is conceptually very simple: It consists in film-
ing the motion of particles and thereafter to reconstruct
their trajectories. However, its practical implementation
is a challenge, and several aspects that must be carefully
considered will be discussed next.
Figure 15.2. Sketch of a typical 3D particle tracking experi-
ment. The central part of the bulk of the flow is illuminated using
two expanded high-power laser beams. Three cameras record
simultaneously the motion of small particle tracers in the bulk.
From Bourgoin et al. [2006].
Resolution Issues For high-resolution measurements,
high-speed cameras with a large number of pixels are
required. As already discussed, three decades of tempo-
ral resolution requires a repetition rate of at least 1 kHz
(assuming large structures evolve with a typical time scale
around 1 s), while four decades of spatial resolution would
in principle require a sensor with at least 10 4
have been generally used [ Voth et al. , 2002; LaPorta
et al. , 2001; Bourgoin et al. , 2006], though alternative and
less expensive solutions using high-power light-emitting
diodes (LEDs) start to be developed [ Del Castello and
Clercx , 2011]. In terms of recording, the stereoscopic
reconstruction requires at least two cameras with two dif-
ferent angles of view to be used simultaneously. In practice
three or more cameras are used. Increasing the num-
ber of cameras has two main advantages: (i) It allows
tracking more particles simultaneously, which is inter-
esting to improve statistical convergence of the mea-
surements, especially when multiparticle problems (for
instance, related to dispersion issues) are investigated, and
(ii) the redundancy for particles which are seen simultane-
ously by more than two cameras improves the accuracy
of the 3D positioning of those particles, thus leading to
an enhanced effective resolution. State-of-the-art optical
Lagrangian systems using three or four high-speed cam-
eras are capable of tracking several hundreds of particles
with 1/10th of pixel of effective resolution. Figure 15.2
shows the three high-speed camera system implemented
by Bourgoin et al. [2006].
10 4 pixels.
State-of-the-art high-speed cameras are typically capable
of recording 10 3
×
10 3 pixel images at several thousands
of frames per second, which yields over three decades in
time and three decades in space. In practice, as discussed
later, experimental noise generally requires to severely
oversample the data, and this lowers the effective time
resolution. On the other hand, several cameras are gener-
ally used simultaneously (as discussed below), which has
the additional benefit of improving the effective resolution
to about 1/10th of a pixel, hence recovering four decades
of effective spatial resolution. The global resolution of a
tracking system is generally a trade-off between tempo-
ral and spatial resolution, as higher repetition rates can
be achieved by reducing the number of pixels and vice
versa. However, due to the impressive progress in high-
speed digital imaging technology, direct optical tracking
has become one of the most accurate techniques in exper-
imental fluid mechanics.
×
3D Issues Complex flows generally involve 3D struc-
tures which require tracking to be done in 3D. This
has two consequences: (i) The flow has to be illumi-
nated in volume (a laser sheet, as done for instance in
PIV, is not sufficient) and (ii) particles must be tracked
in 3D, hence requiring a stereoscopic configuration. In
terms of illumination, as the tracers to be tracked are
generally small (hence the diffused light is dim), the
repetition rate is high (hence exposure time is short),
and the light beam is enlarged (to illuminate a volume),
high-power light sources are required. High power lasers
Data Management Issues High-speed imaging experi-
ments result in a huge data rate. For instance, 1 kHz
acquisitions with three one megapixel sensors recording
at a bit depth of 8 bits represent an effective data rate of
a few gygabytes per second of recording. These usually
require to couple the acquisition system to dedicated data
storage and data processing servers.
The different steps of the data processing, essential for
the optimization of the accuracy of the 3D tracking, are
described next.
 
Search WWH ::




Custom Search