Automation of source-artefact classification

• Main Author: Sebokolodi, Makhuduga Lerato Lydia
• Format: Others
• Language: English
• Published: Rhodes University 2017
• Online Access: http://hdl.handle.net/10962/4920

The high sensitivities of modern radio telescopes will enable the detection of very faint astrophysical sources in the distant Universe. However, these high sensitivities also imply that calibration artefacts, which were below the noise for less sensitive instruments, will emerge above the noise and may limit the dynamic range capabilities of these instruments. Detecting faint emission will require detection thresholds close to the noise and this may cause some of the artefacts to be incorrectly detected as real emission. The current approach is to manually remove the artefacts, or set high detection thresholds in order to avoid them. The former will not be possible given the large quantities of data that these instruments will produce, and the latter results in very shallow and incomplete catalogues. This work uses the negative detection method developed by Serra et al. (2012) to distinguish artefacts from astrophysical emission in radio images. We also present a technique that automates the identification of sources subject to severe direction-dependent (DD) effects and thus allows them to be flagged for DD calibration. The negative detection approach is shown to provide high reliability and high completeness catalogues for simulated data, as well as a JVLA observation of the 3C147 field (Mitra et al., 2015). We also show that our technique correctly identifies sources that require DD calibration for datasets from the KAT-7, LOFAR, JVLA and GMRT instruments.