Next: Displaying the successive civ
Up: A simple image pair:
Previous: Reading the velocity field:
  Contents
To improve the results, come back to the CIV menu, and follow these
successive steps, corresponding to a sequence of operations.
- Select the 'CIV1' operation, so the corresponding parameters
show up. Improve the spatial resolution by selecting smaller correlation
boxes in the civ1 menu, for instance 'ibx,iby' to 19 and
13 (size in image pixels). This is possible because of the good image
quality and high particle density. The use of a smaller box in y allows
to optimize the resolution in this direction. It is also possible
to adjust the search range, using knowledge on extremal velocities,
see histograms displayed by 'uvmat' (on the right side) . We introduce
estimated bounds on each velocity components, 'min' to -2
and 'max' to 6 for 'u' and (-4, 4) for 'v',
and press 'calcul search range' to get values for the search
range 'isx, isy' and a systematic shift. Then we can refined
the grid mesh for the velocity vectors, taking 'dx'=5, 'dy'=5.
- Select the 'FIX1' operation, which will eliminate some false
vectors. Introduce the mask by pressing 'GET_MASK' to suppress
vectors outside the fluid region. It detects the mask image file with
appropriate name 'UV_DEMO/EX_1/frame1_mask_1.png'. If this file
is not present, an interface for making the mask appears. Keep the
default values for the other parameters.
- Select the 'PATCH1' operation, to interpolate the vectors
on a regular grid. Select a high resolution 'nx'=200, 'ny'=60,
representing the number of grid points in which the velocity is interpolated.
These values correspond to dx=dy=5 pixels for an image 1000x300. Choose
the default value 10 for the smoothing parameter 'rho' .
You can later try different values, the smoothing effect increasing
with rho. Keep the default values for the other parameters.
- Select the 'CIV2' operation to improve the correlation results,
using the information on local image deformation, provided by the
previous knowledge on velocity spatial derivatives (calculated in
'patch1'). Use the same grid dx= dy=5 than for civ1. The spatial resolution
can be slightly improved by decreasing the correlation box, using
for instance ibx=iby=15. The search range is determined automatically
using the prior knowledge on velocity, obtained with the civ1 and
patch1 operations. Note that 'civ2' corresponds to a new measurement
from the images, the previous civ1 and patch operations being used
only as an initial guess for the search of optimal correlations.
- Select 'FIX2' and 'PATCH2' with the same parameters
as 'FIX1' and 'PATCH1'.
- Press 'RUN' to run the calculation. The results are stored
in a new subdirectory, A.0, so the previous results are not erased
(you can also set the name of the subdirectory in the 'civ' interface,
by the edit box 'SUBDIR_CIV1'.
The corresponding interface state is shown in Fig. 2. You can directly
get this interface already filled by opening the file 'UV_DEMO/EX_1/frame_civ2.fig',
for instance with the browser of 'uvmat'. Then just modify the input
name and press 'RUN'. The resulting velocity fields are already stored
in 'UV_DEMO/EX_1/VEL.0/frame_1-2.nc', so you can compare with your
results. You can process in several steps, running for instance 'CIV1'
and 'FIX1' alone, then coming back to the civ interface, selecting
only 'PATCH1', 'CIV2'.
The result can be improved again by performing a third civ iteration,
civ3. For that purpose, select only the 'civ2', 'fix2' and 'patch2'
operations with the same parameters as previously, see 'UV_DEMO/EX_1/frame_civ3.fig'.
The previous result is now considered as 'civ1', so set A.0 as the
subdirectory in the edit window 'SUBDIR_CIV1'. Select a
new subdirectory name, for instance 'A.civ3' in the edit window 'SUBDIR_CIV2',
the new result will be obtained there. Further iterations can be similarly
performed, but the improvement becomes negligible.
Figure 2:
The 'civ' interface
|
|
Next: Displaying the successive civ
Up: A simple image pair:
Previous: Reading the velocity field:
  Contents
Joel Sommeria
2004-10-11