Rover Exomars (CNES – since 2004)
Magellium works with the main European actors in stereovision for space exploration, such as CNES, ESA, Thales Alenia Space and Airbus Defence & Space. Our computer vision & robotics team has a proven expertise in 3D measurement, vision-based localisation and system integration.
Self localisation of a rover in an unknown environment is a key and challenging issue in mobile robotics. Magellium has developed for CNES an efficient, stereovision-based method of visual motion estimation (VME).
This image processing technology computes elementary displacements between consecutive frames, and integrates them to localise the rover. Experimental results show good localisation accuracy with a relative error less than 5%, from 122 stereoscopic views acquired on a 27m trajectory.
Magellium is licensed to distribute EDRES, the CNES space exploration robotics development environment SDK. This highly specialised development toolkit provides the complete functional core needed for vision based autonomous motion generation :
- Image acquisition layer,
- Distortion correction layer,
- Stereovision layer,
- Disparity filtering layer,
- DEM reconstruction layer.