Poszukiwanie gwiazd zmiennych w eksperymencie Pi of the Sky

The Pi of the Sky experiment was built to study astrophysical phenomena variable on short time scales (days seconds). The main goal was to search for optical counterparts of Gamma Ray Bursts (GRB), but also for variable stars, novae and optical emissions from Gravitational Wave (GW) sources. Pi of the Sky robotic telescopes are autonomous devices monitoring large fraction of the Sky (each camera observing eld of 200 200) with time resolution of 1 100 seconds and with range of 12m 13m. Dedicated algorithms analyze the collected data in real-time allowing for fast recognition for optical ashes of cosmic origin. LUIZA was designed as a dedicated framework for e cient processing of astronomical images. Data analysis is divided into small, well-de ned steps implemented as the so called processors. The framework allows user to de ne the processor choice and their execution order, as well as all the required parameters at run time, via a simple steering le. The aperture photometry algorithm (dedicated to star identi cation and instrumental brightness determination on a CCD image) adopted from the ASAS experiment was implemented in LUIZA by the author of this thesis. Having the possibility to modify processor parameters the algorithm can be used to process data (CCD images) coming also from other telescopes with di erent exposure times, eld of view or range. The photometry algorithm was used to search for variable stars on Pi of the Sky images. To select variable star candidates a dedicated algorithm based on Multi-Variable Analysis (MVA) methods was prepared. In the rst step the algorithm has to recognize the functional dependiences between di erent parameters describing brightness distributions for signal (vari- able stars) and background (constant stars) based on the simulated variable star samples. These samples were prepared based on single 10s images, were brightness of constant stars was modi ed to model variable stars with di erent shapes of light curves and random period of variability not longer than 3 days. The developed MVA application is trained to recognize variable objects by analyzing shape of their magnitudo distribution. The program is sensitive to both short and long period variable objects.