Stereo vision augmented Medipix3 for real-time material discrimination
1University of Oulu, Faculty of Science, Physics
|Online Access:||PDF Full Text (PDF, 8 MB)|
|Persistent link:|| http://urn.fi/URN:NBN:fi:oulu-202110129129
Oulu : T. Soudunsaari,
|Publish Date:|| 2021-10-12
|Thesis type:||Master's thesis
A secure and sustainable supply of raw materials, such as minerals and metals, is a major challenge for the European Union. Domestic consumption of minerals and metals substantially exceeds production, leading to a high reliance on imports to meet demand, which is driven by the various sectors of the EU’s economy.
This thesis was conducted as a part of the Horizon 2020 funded X-MINE project, which aims to address the issue by combining novel sensing technologies to improve resource characterization and economic feasibility of domestic mining operations. The focus of the thesis is within early stage ore extraction, where a real-time sensing platform can be employed in a way to reduce waste at an early stage, decreasing the environmental footprint generated by downstream processing.
The advances in CMOS technology have enabled the development of photon counting hybrid pixel detectors, such as Medipix3, for noise-free, high resolution X-ray imaging. Combined with high-speed readout electronics, Medipix3 can be used for imaging and identifying higher density intrusions within ore samples in real-time. In addition, developments in GPU-accelerated stereo vision and related hardware have resulted in high-speed stereo vision camera systems. The combination of stereo vision and Medipix3 is explored in this thesis for real-time material discrimination purposes.
In collaboration with multiple European mines, ore samples with reference elemental data were obtained and measured using the combination of stereo vision and Medipix3. In addition, Medipix3 supports a simultaneous dual channel measurement mode, which is used to form a comparison to the performance of the presented system. Measurement principles and physics for both X-ray imaging and stereo vision are discussed. Additionally, calibration methods and algorithms for both measurement modes are presented. Furthermore, methods for data fusion and algorithm performance evaluation are outlined.
Results for both measurement modes are presented, along with the relevant measurement physics. Superior performance is obtained with the augmentation of stereo vision, in part due to adverse effects of high-speed imaging on image quality with X-ray imaging. Dual channel approach requires higher data throughput, which results in a reduced integration time in comparison. Additionally, charge sharing effects due to high resolution reduce the spectral measurement capabilities.
© Tuomas Soudunsaari, 2021. This publication is copyrighted. You may download, display and print it for your own personal use. Commercial use is prohibited.