Changes between Version 168 and Version 169 of UvmatHelp


Ignore:
Timestamp:
Jan 22, 2015, 4:18:08 PM (6 years ago)
Author:
vaillant1p
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • UvmatHelp

    v168 v169  
    772772An alternative possibility is to use the older Fortran program ''CivX'' from the GUI '''civ.fig'''. This can be called directly on the Matlab prompt, by typing  ''>>civ'', or from UVMAT by the menu bar command '''[RUN/PIV (old civ)]'''.
    773773
    774 '''-Modes of frame pair indexing''' A first menu '''[!ListCompareMode]''' selects one among four modes of operation:
     774'''-Modes of frame pair indexing''': A first menu '''[!ListCompareMode]''' on the top left selects one among four modes of operation:
    775775
    776776 * '''PIV''': makes image comparisons on sliding index pairs, as usual for measuring velocities by particle displacements. A second menu '''[!ListPairMode]''' in the panel '''[!PairIndices]''' then selects one among three possibilities (not always available depending on the file indexing):
    777777   * ''pair j1-j2:'' selects a given j index pair (sometimes denoted by letter index) for the whole time series. This is the most common case for PIV. Pair selection is performed in the menu '''[!ListPairCiv1]''' and '''[!ListPairCiv2]''' for the second iteration Civ2, see below. If timing from an XML file [#a3.5Imagedocumentationfiles.xml <!ImaDoc>] has been detected, this is indicated in the edit box '''[!ImaDoc]''' and the corresponding time intervals are indicated (in ms). For some applications, this time interval may evolve in time, so that reference indices ref_i and ref_j are chosen for the display.
    778778   * series (Di): selects a given  index interval for the i index, around a set of reference i indices. The intervals are denoted Di=0|1, -1|1, -1|2, -2|2 ... , corresponding to the index pairs i|i+1, i-1|i+1, i-1|i+2, i-2|i+2 ...around each reference index i. Pair selection is then performed in the menu '''[!ListPairCiv1]''' and '''[!ListPairCiv2]''' like for 'pair j1-j2'.
    779    * series (Dj): same as series (Di) but with the j index.
     779   * series (Dj): same as series (Di) but with the j index. 
    780780 * '''displacement''': compare the current image to a fixed reference frame, as needed to measure the displacement with respect to this reference. The reference index (i index) is set by an edit box '''[num_OriginIndex]'''.
    781781 * '''PIV volume''' performs PIV in volumes for frame sequences with two indices i and j. The second index j is assumed to represent a position in a volume laser scan, while i represents time. Therefore PIV volume requires the selection '''series (Di)''' for '''[!ListPairMode]'''.
     
    791791 * civ1: the initial image correlation process which by itself already provides a velocity field.
    792792 * fix1: detection of 'false' velocity vectors according to different criteria.
    793  * patch1: interpolation and filtering on a regular grid, providing access to spatial derivatives of the velocity (diverrgence, curl, strain).
     793 * patch1: interpolation and filtering on a regular grid, providing access to spatial derivatives of the velocity (divergence, curl, strain).
    794794 * civ2: advanced image correlation algorithm using the result of civ1 as a first approximation.
    795795 * fix2 and patch2: similar as fix1 and patch1, but applied to the civ2 results.
    796796
    797 This series of operations is chosen by selecting the corresponding  check boxes on the left of the GUI '''civ_series''', which give access to the corresponding parameter input panels. Note that the result of each of these operations is stored in the output NetCDF file, so the process can be split in several runs. When an existing NetCDF velocity file has been initially opened, the GUI '''civ.fig''' is automaticaly configured to perform the next operation (fix1, patch1, civ2...) needed in the process.
     797This series of operations is chosen by selecting the corresponding check boxes on the left of the GUI '''civ_series''', which give access to the corresponding parameter input panels. Note that the result of each of these operations is stored in the output NetCDF file, so the process can be split in several runs. When an existing NetCDF velocity file has been initially opened, the GUI '''civ.fig''' is automaticaly configured to perform the next operation (fix1, patch1, civ2...) needed in the process.
    798798
    799799=== 11.2 CIV1: ===
     
    802802  Examples of XML files are provides in /XML_SCHEMAS/ImaDoc_templates.
    803803
    804 ('''[num_!CorrBoxSize_1,_2,_3], [iby])''' set the size (in pixels) of the 'correlation box', the sliding window used to get image correlations. '''[num_!SearchBoxSize_1,_2,_3]''' set the size of the 'search box' in which image correlation is calculated, see figure.  This search box can be shifted with respect to the correlation box by parameters  ([num_!SearchBoxShift_1,_2,_3]). This is useful in the presence of a known mean flow. The default value !SearchBoxSize=(25,25) is generally good, use a larger size for images with few particles, use an elongated box , e.g. (11,41), to optimise the resolution in one direction (for instance in a boundary layer).
    805 
    806 A parameter '''[num_CorrSmooth] ''' controls the smoothing of the image corelation curves used in civ, the default value 1 is generally used.
    807 
    808 The search parameters (!SearchBoxSize,!SearchBoxShift) can be estimated using the press button '''[search range]'''. First introduce the estimated minimum and maximum values of each velocity component u and v. The result depends on the time interval of the image pair. Change the selected image pair if the maximum displacement (ibsx-ibx)/2 is too small (lack of precision) or too large (bad image correlations and risks of false vectors).
     804('''[num_!CorrBoxSize_1,_2,_3])''' set the size (in pixels) of the 'correlation box', the sliding window used to get image correlations. '''[num_!SearchBoxSize_1,_2,_3]''' set the size of the 'search box' in which image correlation is calculated, see figure.  This search box can be shifted with respect to the correlation box by parameters  ([num_!SearchBoxShift_1,_2,_3]). This is useful in the presence of a known mean flow. The default value !SearchBoxSize=(25,25) is generally good, use a larger size for images with few particles, use an elongated box , e.g. (11,41), to optimise the resolution in one direction (for instance in a boundary layer).
     805
     806A parameter '''[num_CorrSmooth]''' controls the smoothing of the image corelation curves used in civ, the default value 1 is generally used.
     807
     808The search parameters (!SearchBoxSize,!SearchBoxShift) can be estimated using the press button '''[search range]'''. First introduce the estimated minimum and maximum values of each velocity component u and v (expressed in pixel displacement). The result depends on the time interval of the image pair.
    809809
    810810[[Image(11-2 CV1.png)]]
    811811
    812 The button '''[TEST]''' allows the user to witness the correlation with a continuous image (in the 'Figure1 Image Correlation' frame which automatically opens) when moving the mouse on the main image. It is even possible to select a point in the image by clicking on it, a new window with the fixed correlation image at this point opens.
    813 The image belows shows the correlation process and the '''!SearchBox''' and '''!CorrBox''' explained before.
     812The button '''[TEST]''' allows the user to witness the correlation as a live plot. It first opens the source image in a new figure '''view_field'''. By moving the mouse in the figure, the local correlation box and the corresponding search box are drawn in the image, and the 2D correlation result then appears in a new figure 'Figure1 Image Correlation' which automatically pops up. It is possible to freeze the current correlation plot by left mouse selection. The figure belows shows the correlation process and the '''!SearchBox''' and '''!CorrBox''' explained before.
    814813
    815814[[Image(civ1_test.jpg)]]
    816815[[Image(Correlation for PIV.png)]]
    817816
    818 The grid determines the positions of measured velocity vectors: it sets the central positions of the correlation boxes (in pixels) for the first image. A default regular grid can be set by the meshes '''[num_Dx] ''' and '''[num_Dy]''' (in pixels). Alternatively a custom [grid->#sec3.6_grid] can be stored in a text file and selected by the check box '''[GET_GRID]'''. This is convenient to limitate the processing to a subregion or to fine tune the resolution.
    819 
    820 A subregion can be alternatively  selected by a mask image, selecting the edit box '''[MASK]'''. If a mask image with an appropriate name is found in the image directory, it wil be detected, and the indication 'xxmask' appears in the edit box. (xx is the number of slices, equal to 1 for a single mask). Otherwise a browser appears to select a (single) mask file.
    821 
    822 Finally thresholds on image intensity can be introduced to suppress underexposed or overexposed parts of the image: select the check box '''[THRESH],''' and edit the boxes '''[!MinIma] ''' and '''[!MaxIma] ''' which then appear.
    823 
    824 The velocity results are stored in the subdirectory set by the edit box '''[subdir_civ1].''' Use the browser''' [list_subdir_civ1]''' to check the existing subdirectories.
    825 
    826 -''' RUN and BATCH: '''
    827 
    828 The PIV calculation is started by the press button [RUN], or [BATCH] if this option has been installed. BATCH sends the same computations as background computing tasks in a network.
    829 
    830 In both cases, the status of the computations can be checked by opening {.cmx } and {.log} files, using the UVMAT browser or any text editor. These files are writtent in the same subdirectory as the NetCDF result files. Each {.cmx} file contains the set of parameters used for a civ computation. It is written by the civ interface just before ordering the computations. By contrast , the {.log} file is produced after completion of the computations, and it contains some information on the process, including possible error messages.
     817The grid determines the positions of measured velocity vectors: it sets the central positions of the correlation boxes (in pixels) for the first image. A default regular grid can be set by the meshes '''[num_Dx] ''' and '''[num_Dy]''' (in pixels). Alternatively a custom [grid->#MaskGrid] can be stored in a text file and selected by the check box '''[GET_GRID]'''. This is convenient to limitate the processing to a subregion or to fine tune the resolution.
     818
     819A subregion can be alternatively selected by a mask image, selecting the edit box '''[Mask]'''. If a mask image with an appropriate name is found in the image directory, it wil be detected, and the indication 'xxmask' appears in the edit box. (xx is the number of slices, equal to 1 for a single mask). Otherwise a browser appears to select a (single) mask file.
     820
     821Finally thresholds on image intensity can be introduced to suppress underexposed or overexposed parts of the image: select the check box '''[!CheckThreshold],''' and edit the boxes '''[num_!MinIma] ''' and '''[num_!MaxIma] ''' which then appear.
    831822
    832823=== 11.3 FIX ===
    833 The FIX operation is used after civ to remove false vectors, using different criteria:
    834 
    835 ''' - {warning flags:} ''' these are flags (vec_F) indicating problems with the image correlation process:
    836  * vec_F(i)=0 or 1 : default , the processing is fine
    837  * vec_F(i)=-2: the maximum of the correlation function is close to the border of the search box (at a distance of two pixels or less), so that its maximum cannot be reliably obtained: the search domain is too small
    838  * vec_F(i)=2: (only in civ1) we keep the less accurate Hart result (resulting from combining correlation functions  from neighborhing points) and  vec_C(i)=-1 is chosen by convention. Reference: Hart D. P. (2000) 'PIV error correction',{ Exp. Fluids} '''29''', 13-22.
    839  * vec_F(i)=3: the optimisation of the correlation function is unstable  or  local Intensity rms of the image =0
    840  * vec_F(i)=4:only in civ2: the difference between the estimator and the result is more than 1 pixel. The two criteria, vec_F(i)=-2, and 3, should be selected (default options). The criterium vec_F=2 (Hart) should not be selected, while vec_F=4 (for civ2) could be selected only for excellent data (otherwise it may be too strict).   
    841 
    842 ''' - {Threshold on the image correlation:}  ''' (vec_C) can be introduced by the edit box [thresh_vecC] (value between 0 and 1). It removes vectors with poor correlation. 
    843 
    844 ''' - {Threshold on the velocity modulus:} ''' (expressed in pixels): it can remove either too small values  (menu '''[inf_sup1]''' set to '<'), or excessive values (menu '''[inf_sup1]''' set to '>'). Erratic zero velocity vectors, produced by a fixed image background, can be eliminated by this criterium. These criteria can be as well applied to the difference with a reference field, defined by a file series (edit box '''[ref_fix1]''') and a field inside the file (menu '''[field_ref1]''').
    845 
    846 ''' - {Mask fix:} '''  It has the same effect as the one used in civ1, but the removed vectors are kept in memory, labelled as false.
     824The FIX operation is used after civ to mark false vectors, using different criteria:
     825
     826''' - warning flags: ''' these are flags (vec_F) indicating problems with the image correlation process.
     827 * vec_F=-2: select to eliminate vectors for which the maximum of the correlation function is close to the border of the search box (at a distance of two pixels or less), so that its maximum cannot be reliably obtained.
     828 * vec_F=3: the optimisation of the correlation function is unstable  or  local Intensity rms of the image =0. Must be selected.
     829 
     830''' - Threshold on the image correlation:  ''' (vec_C) can be introduced by the edit box '''[num_!MinCorr]''' (value between 0 and 1). It removes vectors with poor correlation. 
     831
     832''' - Threshold on the velocity modulus: ''' (expressed in pixels). It can remove either excessive values (threshold set by '''[num_MaxVel]''') or too small values (threshold set by '''[num_MinVel]'''). Erratic zero velocity vectors, produced by a fixed image background, can be eliminated by the latter criterium.
    847833
    848834''' - {Manual fix:} ''' Interactive fixing with the mouse is also possible, see [section 4.2->#4.2].
     
    855841
    856842=== 11.4 PATCH ===
    857 -'''PATCH1: ''' interpolates the velocity values on a regular grid with a smoothing parameter rho. This also provides the spatial derivatives (vorticity, divergence) needed for the refined processing civ2. The numbers of grid points in x and y are set by nx and ny. The vectors which are too far from the smoothed field (erratic vectors) can be eliminated, using a threshold expressed in pixel displacement. The thin plate spline method must be done by subdomains for computational saving, use the default value 300.
     843'''PATCH1: ''' interpolates the velocity values on a regular grid with a smoothing parameter rho. This also provides the spatial derivatives (vorticity, divergence) needed for the refined processing civ2. The numbers of grid points in x and y are set by nx and ny. The vectors which are too far from the smoothed field (erratic vectors) can be eliminated, using a threshold expressed in pixel displacement. The thin plate spline method must be done by subdomains for computational saving, use the default value 300.
    858844
    859845=== 11.5 CIV2 ===
    860846-'''CIV2:''' provides a refined calculation of the velocity field, using the civ1 result as previous estimate. The search range is determined by the civ1 field so there is no parameter. Use the option decimal shift to reduce peaklocking effects (advised). The option deformation is useful in the presence of strong shear or vorticity. Then image deformation and rotation is introduced before calculating the correlations. The other parameters (rho, ibx,iby, grid, mask) are used like for civ1 (take the same by default). The image pair for civ2 can be different than the one used in civ1 to get the first estimate. It is generally advised to use a moderate time interval for civ1, to avoid false vectors, and to take a larger intervel for civ2, in order to optimize precision. As civ2 already knows where to look and takes into account image strain and rotation (with the option deformation) it allows for higher time intervals.
    861847
    862 -'''FIX2:''' like FIX1, except for the different flags vec_F provided by civ2. Use the default options.
     848-'''FIX2:''' like FIX1, except for a new possible flag value vec_F=4 which indicates that the difference between the estimator and the result is more than 1 pixel. This should be selected for excellent data (otherwise it may be too strict).   
     849
    863850
    864851-'''PATCH2:''' like PATCH1