DigImage: Particle Tracking
:::: IMAGE CAPTURE
:::: PARTICLE LOCATION
:::: PARTICLE MATCHING
:::: STRUCTURE OF FILES
:::: BASIC OPERATION OF TRK2DVEL
:::: CONFIGURATION OF TRK2DVEL
:::: PARTICLE-TAGGED DATA
:::: IMAGE PREPROCESSING
Trk2DVel should not normally be entered within a given directory until after at least one experiment has been tracked (or copied to) that directory. Trk2DVel expects the file Track2D.DIG (used to store the controlling parameters for particle tracking) to exist in the directory. If it does not, Trk2DVel will generate an error message to this effect.
In the same manner as DigImage generates the status file Status.DIG to store the operational status of DigImage, Trk2DVel creates the status file Trk2DVel.DIG. The first time Trk2DVel is started from a given directory it will be unable to find the Trk2DVel.DIG file and will generate the message
WARNING: Using default path settings - use [;C: Configure Trk2DVel] to change Press any key to continue...
This warning message indicates one of the primary functions of Trk2DVel.DIG will use default values. In particular Trk2DVel.DIG stores information on how the particle path information is to be used to evaluate the particle velocities. Trk2DVel can be configured to utilise the path information in a number of different ways using [;C: Configure Trk2DVel], the chosen configuration then being stored in Trk2DVel.DIG along with other status information. The configuration menu has the following options:
Configure Trk2DVel A Minimum acceptable path length : C Use large particles : D Use small particles : E Use elliptical particles : F Use faint particles : I: Invalid samples - restart criteria P Use first matching in a path : R: Local intensity regions : T Use particles matched two samples back: U: Type of calculation for velocity : V Period for calculating velocity : Z Set command file search path Q Return to parent menu ->
Central to the evaluation of velocity from the particle paths are options [;CA Minimum acceptable path length], [;CU: Type of calculation for velocity] and [;CV Period for calculating velocity]. The first of these specifies the minimum length a path must be before it may be considered valid, while the third gives the length within this path to be used for calculating the velocity. The minimum length should always be greater than or equal to the velocity period; in most cases the two will be set equal. Choosing the period for the velocity is a compromise between velocity resolution, the velocity time scales and the average length of time a particle remains in the light sheet. The larger the number of time steps which may be used to calculate the velocity the more accurately the velocity may be calculated, provided the velocity has not changed significantly during that period. The period should always be more than two time steps to ensure invalid new particle matchings do not contaminate the results. (Note: times may be specified in the minutes:seconds and ::fields format of DigImage, and as ;samples giving a period of samples*time_step). Typically a velocity period of four or five time steps (i.e. ;4 or ;5) provides an acceptable balance between the various criteria. As a check the same paths may be processed using different velocity periods and their results compared.
The menu produced by [U: Type of calculation for velocity] determines how the positions of the particles falling within the velocity period are used to determine the velocity for the mid-point of the path:
Type of Calculation for Velocity C Chebeshev linear fit L Linear least squares M Quadratic least squares P Extreme points Q Return to parent menu ->
With the exception of [P Extreme points], all these options use information about the particle positions throughout the [V Period for calculating velocity] ([P Extreme points] simply performs a finite difference calculation on the extremes of this period). In general the least squares approach is to be preferred with [L Linear least squares] evaluating linear fits to the x t and y t data within the velocity period, while [M Quadratic least squares] fits a quadratic polynomial to the same data. For good quality data where interlace filtering has not been necessary, there is little to choose between the linear and quadratic fits to the same number of time steps. However, if the average path length is sufficiently long, then the quadratic fit may be used over longer periods, while the linear fit is less prone to noise with lower quality data. A comparison of the velocity probability density functions for the two approaches generally highlights any differences.
In addition to determining how the particle data is used to determine the velocities, the [;C: Configure Trk2DVel] menu also specifies which particles and paths may be used. Options [;CC Use large particles], [;CD Use small particles], [;CE Use elliptical particles] and [;CF Use faint particles] may be used to deselect certain categories of particles such that a path will be considered to terminate when the particle falls in one of the deselected categories. For example, if [;CF Use faint particles] is set to No, and on a path twenty time steps long the particle image at the fifth time step on a path was characterised as faint (i.e. it satisfied only the lower band of thresholds during the tracking phase), then the path would be considered as two paths, one extending up to the fourth time step, and the other from the sixth to twentieth time steps. In general the velocity structure of a well set up experiment should not be sensitive to the settings of these options. They will affect the number of particle paths extending over at least a given number of time steps, but should not affect the measured velocity probability distribution.
While tracking, DigImage performs two stages of matching. The first is between the new time step and the previous time step, and the second between those particles not matched in the first stage and those not matched two time steps back. Depending on the particle density, there may be around 10% additional matches in this second stage. The option [;CT Use particles matched two samples back] allows paths including such a matching to be deselected. If such paths are included, then a simple linear interpolation is used to estimate the position of the particle at the time step where it was not located.
In rare cases when using low quality data (typically data tracked with [;USPM Maximum matching distance] set too large) it may be desirable to exclude the first matching in a path. This may be achieved through [;CP Use first matching in a path].
During the particle tracking DigImage records a number of events which may affect the quality of the tracking data. Subsequently during the analysis phase, it may be desirable to reject data associated with these events by stopping all the particle paths when one of these events occurs. The menu [;CI: Invalid samples - restart criteria] determines which events force the particle paths to be restarted. The events concern either the root mean square error in the reference point mapping (a large rms error indicates a low quality image), or the time code pulses placed on one of the audio channels (used to trap errors in the VTR tape position). The options in the menu are:
Invalid Sample Restart Criteria R Retry for reference point rms error : S Reference point rms error not satisfied : T Unable to find time code pulses : U Time code pulse error > 2 : V Time code pulse error < -2 : Z Time code pulses re-zeroed : Q Return to parent menu ->
The first two options involve the rms error in the reference mapping. [;CIR Retry for reference point rms error] is not normally a serious event. If during the tracking phase DigImage detects an rms error larger than the limit set by [;USZ Limit on rms error for mapping] then it will try to improve the quality of the image by repeating the acquisition of the frame. In general the quality will be improved on the first retry and there is no need to restart the particle paths during the analysis. However, if after five retries the rms error is still too large, then DigImage will continue tracking, noting that it could not satisfy the requirement. Under most circumstances samples flagged with this condition will be of unacceptable quality, and so the particle paths should be restarted. This is achieved by setting [;CIS Reference point rms error not satisfied] accordingly.
The other four options in this menu concern the time code pulses on the audio channel. Time code pulses must be added to the video tape by the DigImage option [;VVA Add time code pulses] prior to tracking the particles. The time code pulses are used to detect and correct any errors in the tape transport due to the limitations of the video hardware. If the time code pulses are not present, or there is something wrong with them, then DigImage may not be able to reconstruct the time sequence with the required 100% accuracy. [;CIT Unable to find time code pulses] determines whether or not the particle paths are to be restarted if DigImage was unable to locate the pulses while tracking the particles normally this event should restart the particle paths.
The time code pulses are designed to correct errors of between +4 and -2 video fields (+2 and 1 frames). However, under normal operation the errors should be confined between +2 and 2 fields. An error detected outside this range normally indicates a more serious timing problem, and so should be used to restart the particle paths. Similarly under some (very rare) circumstances the error is such that DigImage is not sure whether the tape is behind or ahead the position it thinks it should be in. The strategy here is that DigImage will re-zero the time code mechanism, effectively forcing the current situation to be correct. [;CIZ Time code pulses rezeroed] should normally be set to restart the particle paths should this event be detected.
Another option in [;C Configure Trk2DVel] is [;CR: Local intensity regions] which produces a menu controlling the use and meaning of the different local intensities saved along with the particles under some combinations of the controlling tracking parameters.
Background Intensity I Included regions : 0 Concentration for region 0 : 1 Concentration for region 1 : 2 Concentration for region 2 : 3 Concentration for region 3 : 4 Concentration for region 4 : 5 Concentration for region 5 : 6 Concentration for region 6 : 7 Concentration for region 7 : Q Return to parent menu ->
[;CRI Included regions] determines which of the intensity regions are to be included in the paths used. If a path contains a particle image with a local background intensity falling into one of the levels which is not included using this option, then the path will be considered as stopping before the corresponding time step (and possibly a new path starting at the next time step).
Some of the facilities in Trk2DVel are able to operate either on velocity or concentration flux. For concentration flux measurements, options [;CR0 Concentration for region 0] to [;CR7 Concentration for region 7] should be set up with the correspondence between the eight intensity regions and the concentration of some scalar field. The required calibration procedure is beyond the scope of this document.
The final option in [;C
Configure Trk2DVel] is [;CZ
Set command file search path]. This option
allows the specification of an optional search path for command
files when started by the !P
directive (either interactively or from within another command
file). If a requested command file is not found in the current
directory, then the path specified here will be searched. If the
file is still not found, the %DigImage%\Macros
directory will be searched.
Goto next document (Facilities)
DigImage documentation page
DigImage home page
Stuart Dalziel's home page