Stevanovic, Uros

PhD thesis, Faculty of Electrical Engineering and Information Technology, Karlsruhe Institute of Technology, 2017.


This dissertation proposes a novel smart camera platform serving as a flexible data acquisition system for scientific applications. Current technological progress offers increasing performance in the areas we consider, namely high data-throughput, data processing, and detector performance. Prevalent data acquisition solutions typically focus on one of these aspects. However, driven by science, experiments experience increasing demands in terms of data throughput, speed and flexibility. In this dissertation, we introduce a system which, in addition to being able to provide high-speed data transfer, is also capable of interpreting the incoming information at an early stage. In order to demonstrate the full potential of the smart camera platform, we focus on X-ray imaging with synchrotron light sources. X-ray imaging applications can investigate the traits of technological and biological processes over microseconds for radiography, and milliseconds for tomography applications. These applications may require different sensors, and include complex experiment operations. The new smart camera platform is part of a larger project, UFO, which introduces a new concept for X-ray imaging. On-line data assessment is used to provide a data-driven feedback and active management of both the process and data acquisition procedure. This is accomplished using a GPU platform for fast reconstruction, embedded on-camera data processing, and integrating smart camera in a high-throughput data acquisition system. The final design of the smart camera platform consists of a custom high-performance FPGA board, providing continuous data transfer, embedded image processing, and a flexible input stage. In the IMAGE beamline of ANKA, camera is integrated in the new control system, and used in real-life applications. A maximum data-throughput of up to 8 GB/s is achieved. A custom image-based algorithm is implemented in the FPGA, with stringent real-time requirements, able to increase native sensor speed up to five times while reducing the amount of transfered data. Several image sensors are used, with resolutions of up to 20 megapixels and frame rates of up to 5 kfps. The smart camera platform was also used in non-imaging applications, stemming from the flexible input stage. The proposed camera architecture enables the user to modify the current system for any kind of high data-throughput applications, and to modify and implement custom processing algorithms.


First assessor: Prof. Dr. M. Weber
Second assessor: Prof. Dr.-Ing. Dr. h.c. J. Becker