zur Startseite

Research Focus

The research of the department of Parallel Systems focuses on high performance hardware/software systems. In the following, a summary of the research activities and how they are related to their application domain of imaging and sensor systems are presented:

 

High Performance Hardware/Software Systems for

Real-Time Imaging and Sensor Systems

 

The real-time processing of data captured by high-speed 2D or 3D imaging and sensor systems benefits from dedicated parallel hardware architectures with suited parallelized algorithms forming high performance hardware/software systems. For intelligent sensor systems the hardware and software is integrated into the sensor system. An important class of such hardware/software systems are systems based on reconfigurable computing using field programmable gate arrays (FPGA). In order to obtain high performance with this approach, a good match between the architecture and algorithm is required which is suited for the fine grain parallelism of FPGAs. An example of a high-speed 2D imaging sensor system with an integrated and dedicated FPGA-based hardware architecture for real-time object analysis shown in Fig 1.

 

For high performance volume data processing, computed tomography (CT) - see Fig. 2 - is a good example for which the high performance hardware/software system is based on a general purpose computer with an additional graphics processing unit (GPU) as hardware accelerator and parallel software written in OpenCL. Due to the increasing performance of GPU accelerators over the last years, advanced and compute-intensive iterative CT reconstruction algorithms become more and more applicable for 3D computed tomography with respect to their acceptable computing time on modern CPUs or GPU-clusters. Even in the case of high-resolution 2D CT-detectors above 10Mpixel the computing time of iterative reconstruction algorithms may be considered as acceptable. 

Computed tomography generally known from noninvasive examination in medicine or biology - see Fig. 3 - is additionally and increasingly used in non-medical disciplines like materials science and mechanical engineering for the extraction of geometry models of the respective objects, see Fig 2. These geometry models can be applied to scientific numerical simulations to derive physical properties of these objects like e.g. the fluid flow through porous media. This concept to derive physical properties by numerical simulation of extracted geometry models from CT volume data has been transferred here to the physical design domain of high-speed digital circuits and RF circuits. For example S-parameters and the line impedance of a stripline or microstripline can be derived from the CT scan of an IC package. This approach can also be used for the physical design of high performance hardware/software systems. Exemplarily, several selected publications about hardware or software aspects of the high performance hardware/software systems are listed as follows:

  1. M. Najmabadi, T.-H. Tran, S. Eissa, H. S. Tungal, S. Simon, An Architecture for Asymmetric Numeral Systems Entropy Decoder - A Comparison with a Canonical Huffman Decoder, Journal of Signal Processing Systems, accepted, 2018
  2. Y. Baroud, J.M.M. Velarde, Z. Wang, S. Kieß, M. Najmabadi, J. Guhathakurta, S. Simon, Architecture for parallel marker-free variable length streams decoding, Journal of Real-Time Image Processing, 1-20, 2017
  3. M. Klaiber, D.G. Bailey, Y. Baroud, S. Simon, A Resource-Efficient Hardware Architecture for Connected Components Analysis, IEEE Transactions on Circuits and Systems for Video Technology, 2016, 26 (7), 1334-1349 13.    
  4. M. Klaiber, D. Bailey, S. Simon, A Single-Cycle Parallel Multi-Slice Connected Components Analysis Hardware Architecture, Journal of Real-Time Processing, 1-11, 2016
  5. J. Hillebrand, S. Kieß, S. Simon, Signal Integrity Model Extraction based on Computed Tomography Scans - Analysis of the Required Voxel Resolution, IEEE Transactions on Electromagnetic  Compatibility, 2015

The high performance hardware/software systems are realized as proof-of-concept to analyze and demonstrate the impact of the investigated parallel architectures and algorithms. As the performance and efficiency of dedicated hardware/software systems are application dependent, a suited application domain for the analysis and demonstration of the high performance hardware/software systems has to be chosen. The selected application domain is that of real-time imaging and sensor systems which are used in various scientific disciplines like natural science, materials science, simulation science, process or civil engineering. These scientific disciplines benefit from those demonstrators of the high performance hardware/software systems as novel imaging and sensors systems with unique properties become available. One example is an imaging system with integrated real-time processing for metrology with image data streams of 10.000fps or more, see Fig. 1.

Due to this focus, the considered hardware/software systems are realized often in interdisciplinary funded projects in cooperation with institutes of the chosen scientific discipline. Exemplary, several high performance hardware/software system demonstrators for different disciplines are subject of the following publications. They are mostly written as joined publications with the cooperating institutes. They typically summarize the impact of the respective high performance hardware/software system on the scientific discipline as result of the interdisciplinary research.

  1. M. Klaiber, Z. Wang, S. Simon, A Real-Time Process Analysis System for the Simultaneous Acquisition of Spray Characteristics, Buch Process-Spray, Springer Verlag, p. 265-305, 2016
  2. J. Guhathakurta, D. Schurr, G. Rinke, R. Dittmeyer, S.Simon, Simultaneous In Situ Characterisation of Bubble Dynamics and a Spatially Resolved Concentration Profile: A Combined Mach–Zehnder Holography and Confocal Raman-Spectroscopy Sensor System, Journal of Sensors and Sensor Systems 6 (1), 2017
  3. J. Stark, T. Rothe, S. Kieß, S. Simon, A. Kienle, Light scattering microscopy measurements of single nuclei compared with GPU-accelerated FDTD simulations, Journal: Physics in Medicine and Biology 61 (7), 2749-2761, 2016
  4. M. Zimmermann, A. Tausendfreund, S. Patzelt, G. Goch, S. Simon, et al., In-process Measuring Procedure for Sub-100 nm Structures, Journal of Laser Applications, 2012
  5. G. Knizia, W. Li, S. Simon, H.-J. Werner, Determining the Numerical Stability of Quantum Chemistry Algorithms, Journal of Chemical Theory and Computation, 2011