Difference between revisions of "Projects:Automatic generation of optimal marker patterns"
(→Contact) |
(→Estimated level of difficulty) |
||
| Line 21: | Line 21: | ||
== Estimated level of difficulty == | == Estimated level of difficulty == | ||
| − | + | Probably an elaborate problem (would make a good Master's thesis). | |
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
== Provided data == | == Provided data == | ||
Revision as of 09:42, 21 May 2019
Contents
Overview
The Vicon Nexus software automatically identifies patterns of markers within the 3D point cloud of marker positions which is being tracked. To distinguish individuals, it is necessary to give a unique pattern to each individual, which the software can reliably detect and tell apart from the others. The key goal of this project is to establish an automated process to generate optimal versions of these patterns for different types of experiments.
Contact
- Mate Nagy, mnagy@orn.mpg.de
- Hemal Naik, hnaik@orn.mpg.de
- Nora Carlson, ncarlson@orn.mpg.de
Aims
- Generate unique patterns which can be optimally tracked and distinguished given the input design criteria (e.g. maximum number of markers, marker size, dimension of base plate, minimum distance between points).
- In particular, controlled experiments need to be performed to find out what works well for the Vicon software systems, and what is required of a pattern for tracking to work well.
- Use VICON calibration to actually create a full 36 camera setup, simulate performance of these designs in terms of expected detection accuracy and performance in different areas of the barn.
- To reduce amount of manual work, it would be desirable to automatically produce precise tags (using CNC, laser cutting, 3D printing ...) based on a given pattern design.
Estimated level of difficulty
Probably an elaborate problem (would make a good Master's thesis).
Provided data
See Vicon:Data format documentation.
This project requires a lot of experimentation with different marker setups and a number of controlled experiments, which need to be designed by whoever attacks it.
Suggested approaches
Estimate 2D invariant values for various designs using theory of cross ratios and suggest creating individual patterns. Simulation of the data can be done using 1 or 2 virtual cameras.
