http://arxiv.org/abs/1911.11826
Point Sources (PS) are one of the main contaminants to the recovery of Cosmic Microwave Background (CMB) signal at small scales, and their careful detection will be important for the next generation of CMB experiments like CORE. We want to develop a method based on Fully Convolutional Networks (FCN) to detect PS in realistic simulations, the Point Source Image Detection Network (PoSeIDoN), and compare its performance against one of the most used PS detection method in this context, the Mexican Hat wavelet 2 (MHW2). The frequencies for our analysis are the 143, 217 and 353 GHz Planck channels. We produce PS realistic simulations at each frequency including possible contaminating signals as CMB, Cosmic Infrared Background, Galactic thermal emission and instrumental noise. We first produce a set of training simulations at 217 GHz. Then we apply both the trained FCN and the MHW2 to recover the PS in the validating simulations at all the frequencies, comparing the results by estimating the reliability, completeness and flux density estimation accuracy. In the extragalactic region with a 30 deg galactic cut, the FCN successfully recover PS with 90% of completeness corresponding to 300, 139 and 227 mJy for 143, 217 and 353 GHz. On the same validation simulations, the MHW2 with a 4sigma flux density detection limit, recover PS till 298, 173 and 227 mJy at the 90% completeness. In all cases the FCN produce a much lower number of spurious sources with respect the MHW2. As expected, the results on spurious sources for both techniques worsen when increasing the frequency or reducing the galactic cut to 10 deg. Our results suggests that FCN are a very promising approach to detect PS using data from CMB experiments, providing overall better results with respect to the more usual filtering approaches.
L. Bonavera, S. Gomez, J. González-Nuevo, et. al.
Thu, 28 Nov 19
64/70
Comments: 9 pages, 4 figures, submitted to Astronomy & Astrophysics
You must be logged in to post a comment.