Fiducial Marker-Based Multiple Camera Localisation System

Assigment:

1. Learn principles of multi-camera geometry.
2. Learn about fiducial markers used in the mobile robotics.
3. Propose a set of key performance criteria to assess the markers' performance in relevant mobile robotics scenarios,
where accurate 6DOF estimation over large-scale areas is required.
4. Evaluate the performance of the selected markers in a single-camera configuration.
5. Choose an appropriate marker detection method and design and implement its extension for multi-camera configurations.
6. Using the key performance criteria set in (3), evaluate the impact of the implemented extension.

‍

Abstract:

In this thesis, the multi-camera localisation system based on fiducial markers is presented together with a modified algorithm to detect black-and-white circular fiducials. The introduced localisation system originates from the real-time single-camera fiducial marker system with an
estimation of the position and orientation in the 3D space and a distinguishable encoded variable identification. This multi-camera localisation system built on cheap and widely available web cameras represents a publicly available and open localisation system as an alternative to the currently available expensive, closed systems using high-end cameras and specialised hardware requiring tedious deployment. In order to assess the performance, the localisation system is tested against the single-camera original method widely used in the fields of mobile and swarm robotics. We created simulated testing environments allowing dynamically changeable simulated scenarios, and we also collected a real-world dataset of an application on the mobile robot external localisation. An automatic evaluation and simulation framework was introduced to make the testing process more manageable.

Result:

The thesis is for download on CVUT's dspace.

‍