PBI - the Particle Bureau of Investigation
==========================================

PBI is a toolbox for 3D-PTV (Particle Tracking Velocimetry). It incorporates a 
number of graphical tools with a set of command-line scripts for performing the 
different stages of a 3D-PTV analysis.

The package has been separated from Yosef Meller's PhD repository, and therefore 
it may not be as well organized as your ideal package (yet), but it's still quite 
useful.

The package relies on OpenPTV for the heavy machinery, and most of the relevant 
documentation and tests are in OpenPTV. There is always some documentation 
available (at least for the command-line tools) by running them with the '--help' 
option. E.g.

  > ptv/dumbbell.py --help

All scripts work with YAML configuration files. I should add some examples, right? 
hopefully I'll do so soon, in the meantime just ask me.


Tools in the package
--------------------

Calibration can be done mainly with these tools:

 * ui/cam_calib.py is a graphical tool similar to PyPTV except very different. It 
   does one camera at a time, and doesn't save any results unless asked to do so.
   It allows you to change parameters on the fly, with spin boxes, without editing
   text files.
 
 * evolution/evo_cal.py - starts with a relatively broad range of possible 
   parameters and evolves a global solution which is quite good. Depending on how 
   much you want to wait, of course.

To check your calibration:

 * ptv/view_calib.py - a graphical tool using TraitsUI [1] and Tracer [2] that
   gives you a visualization of how good your calibration is by shooting rays from 
   the camera through detected points (after distortions). You need the rays to
   converge on the balls that represent known calibration points.
 
 * ui/epi_checker.py - a tool similar to the PyPTV main screen. It shows 4 cameras
   together and alows you to mark epipolar lines with minimum hassle.

[1] http://docs.enthought.com/traitsui/
[2] http://yosefm.github.io/tracer/

There are supporting tools for calibration with different methods.

Dumbbell:

 * ptv/dumbbell.py - does detection of dumbbell from initial user marking,
   just like the old Matlab tool only faster and more configurable.

 * ptv/dumbbell_correct.py - improves an existing calibration using the results of 
   ptv/dumbbell.py

Multiplane:

 * ptv/gen_multiplane.py - generates configuration files and shifted point files 
   for a multiplane calibration. Might not be entirely up to date on the config 
   format, be careful. Note that it generates several auxiliary files, but you
   have full control over where they end up, so use it and don't clobber your own
   files.

 * ptv/gen_multi_init.py - Generates an initial guess for multiplane calibration 
   that is hopefully different than each of the single-plane calibrations.

 * ptv/multiplane.py - does the actual multiplane calibration using outputs of the 
   other tools.

And also,

 * ptv/shake.py - calibration by the shaking method.

Oh, and the evolution/ directory has two more evolution scripts that attempt 
to solve some form of a multi-camera calibration, either by maximizing 
correspondences or by comparing points in 3D. I have not touched them in a
very long time and do not trust them to work. But I give the code, should 
anyone care to look into it in the future.

The actual PTV process for a calibrated scene is served by these tools:

 * ptv/sequence.py - what you think. Note the outputs are configurable,
   so you don't have to clobber your results every time. Depends on an 
   external parallel processing package [3]

[3] https://github.com/yosefm/parallel_runner

 * ptv/track.py - the forward tracking pass. I have not yet bound the backward
   tracking to Python, but also have never seen it do anything good so why bother.

That's pretty much it. All post-processing you should do with Flowtracks, which is 
a different package.