CrynodebMaterials Science has in recent years become a high priority research area, having been identified as a growth sector for the UK economy over the next decade. Breakthroughs in this field are likely to have a significant impact on every area of our lives.
There has recently been a trend toward automation at beamlines which is driven by rapid technology advancement. This technology advancement has improved the quality of experiment data and has allowed data collection times to improve exponentially.
The Materials Science Research Group in the Institute of Mathematics, Physics and Computer Science, at Aberystwyth University have achieved international recognition for their research on materials under extreme conditions. They have a rich history in the development and use of specialist instruments to conduct real time surface analysis. Their custom made instrumentation has allowed them to greatly improve experiment throughput. Automation of the group's ultra high vacuum chambers is therefore a further enhancement that is advantageous, important, necessary and inevitable.
This thesis presents the research undertaken to study what is required to provide automated sample positioning inside vacuum chambers that are operated under ultra high vacuum conditions, as the first step towards automation. As part of the research, a prototype automated positioning system that employs state of the art model based visual tracking techniques was developed to gain an understanding of the challenges the ultra high vacuum environment presents.
Experimentation was carried out to assess the effects of different lighting conditions on tracking, evaluate the tracking library, extract suitable extrinsic parameters for tracker initialisation, and evaluate both monocular and stereo mode tracking.
Key findings were that the model based tracking is a suitable approach for an automated positioning system but that performance depends on having suitable port placement for the cameras. Stereo tracking provided the best performance but was still prone to divergence at certain relative positions of the manipulator. On linear runs the average error was 0.06mm. On rotational runs, anti-clockwise runs proved better with an average error of 2o to 3o.
The high errors of mixed rotational and linear tracking runs did not match the visual outputs indicating that there were inherent errors in the data evaluation. Tracking output video footage is available at . More work is needed to take the system forward and close the tracking loop. Recommendations for improvements were provided.
|Andrew Evans (Goruchwylydd) & Frédéric Labrosse (Goruchwylydd)