A Statistical Approach for the Recognition of Extended Radio Sources in Large-Area Sky Surveys

• Activity Number: 655
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Martin Silerio-Vazquez*+ and Heinz Andernach and Carlos Rodríguez and Johan Van Horebeek
• Companies: Iowa State University and University of Guanajuato and University of Guanajuato and Center for Research in Mathematics
• Keywords: Astrostatistics ; radio sky surveys ; morphological feature extraction ; classification ; logistic regression

Abstract

With the aim to identify and recognize giant radio galaxies in surveys that cover large parts of the sky at radio frequencies, we designed a new algorithm that works directly on the intensity images in FITS format. We apply this algorithm to the 2326 images of the NRAO VLA Sky Survey covering all the sky north of declination -40 degrees (82 per cent of the sky). We extract a list of candidate emission regions whose radio surface brightness is above 2.5 times the noise level in a contiguous area containing at least two sources listed in the NVSS catalog, i.e. the regions generally exceed about four antenna beam sizes. After an appropriate preprocessing, we identify a few morphological features and symmetry measures which turn out to have a discriminative power to contain potential giant radio galaxies. Using these features, a classifier based on logistic regression is estimated. Finally, the classifier's potential is validated using a set of emission regions that are visually classified by astronomers.