Editing Projects:Automatic optimization of VICON system and recording parameters

== Overview == 

The aim is to find the optimal setting of the parameters of the Vicon system
based on the problem which needs to be solved. For example, if the animals move very fast or slow, or if their motion is mainly on the
surface vs. using the full 3D, or based on the different marker sizes and pattern designs. 

Currently, the system has 3 different types of parameters:

# things that needs to be set by adjusting the apparatus (focusing and aperture of the cameras, their directions, etc) (Note: we may not want to change this)
# parameters that can be set digitally from the Vicon software before recording and that cannot be changed later (store intensity, parameters for resolving overlapping blobs, etc.)
# parameters that can be set digitally and affect the Vicon's software tracking that can be change later to re-run the analysis pipeline

In both category (2) and (3) there are around 5-10 important parameters. Finding the optimal setting is very hard doing this manually. Solving challenge (3) could be done offline, so it is an easier task. (2) requires actual measurements there. We have a robot vacuum cleaner that goes around in the full area, and can be used for testing. So if there is an algorithm that sets different parameters in the software and robots goes around after the algorithm evaluates the tracking output and changes the parameters until it finds an optimal setting.

== Contact ==

* Mate Nagy, mnagy@orn.mpg.de

== Aims ==

A systematic approach to finding the optimal settings of the system parameters, given the tracking problem to be solved. The more automation the better.

== Estimated level of difficulty ==

This seems like an elaborate problem, which would make an interesting Master's thesis (or three).

== Provided data ==

See [[Vicon:Data format documentation]].

This project likely requires a lot of experimentation with the different settings of the Vicon system and a number of controlled experiments, which need to be designed by whoever attacks it.

== Suggested approaches ==

Suggested things which could be tried out on subproject (1):

* Input :Use the .xcp and image error , world error criteria to judge calibration quality.
* Output: Estimate the cameras or areas that would perform poorly due to bias in calibration.