Publications Repository - Helmholtz-Zentrum Dresden-Rossendorf1 Publication
Real-time Object Recognition for Ultrafast Electron Beam X-ray Computed Tomography
Ultrafast electron beam X-ray computed tomography  (UFXCT) is a non-invasive imaging technique based on scanning an electron beam on a tungsten target. This way, a moving X-ray source is generated without mechanically moving parts allowing for very high imaging rates up to 8000 fps.
This technique is used, e.g., for the investigation of multiphase flows, such as bubbly flow in industrial bubble column reactors. The goal of the ROOF experiment is to investigate the hydrodynamics of such a bubbly flow by apply UFXCT for scanning and tracking of objects alongside a vertical axis in real time based on the acquired cross-sectional images. To accomplish the tracking, software is needed to recognize objects in the reconstructed images, as a human would not fulfill realtime constraints. The current CPU-based implementation is the slowest step in the current workflow, thus, the goal of this work is to design a faster algorithm. In this work, the approach used by the RISA  (Realtime Image Stream Algorithms) software to realize the tracking of objects is presented and improved by using the GPU to recognize the objects.
- Data publication: Real-time Object Recognition for … (Id 34999) HZDR-primary research data are used by this (Id 34952) publication
TU Dresden, 2022
Mentor: Dominic Windisch