Supporting dusty plasma experiments in massive parallel environment

OData support
Supervisor:
Reichardt András László
Department of Broadband Infocommunications and Electromagnetic Theory

In my thesis I show the apparatus of the dusty plasma experiment.

During the experiment the particles form a crystalline structure and image sequences are taken of them with a high speed camera.

Processing these images results in the particle's positions and statistical distributions of the system.

I describe the detection procedure of the particles using filtering and adaptive decision level determination.

Instead of using the common and trivial FIR Gauss filter I propose and implement the more efficient median filter.

For computing the particle's position I implement the momentum method,

which is computationally more demanding but it provides sub-pixel resolution.

I give an overview of the OpenCL framework, which is used for parallel programming.

Applying the above introduced framework and taking into account the properties of the devices available I have composed the

program's principle steps.

These steps are optimized to achieve fast run times.

I benchmark the program's performance on CPU, on GPU and on many integrated core (MIC) card.

Finally I combine the program with the image acquiring software of the high speed camera for online image processing.

Downloads

Please sign in to download the files of this thesis.