Point source detection and extraction from simulated Planck TOD using optimal adaptive filters

Abstract

Wavelet-related techniques have proven useful in the processing and analysis of one and two dimensional data sets (spectra in the former case, images in the latter). In this work we apply adaptive filters, introduced in a previous work (Sanz et al. 2001), to optimize the detection and extraction of point sources from a one-dimensional array of time-ordered data such as the one that will be produced by the future 30 GHz LFI28 channel of the ESA Planck mission. At a $4σ$ detection level 224 sources over a flux of 0.88 Jy are detected with a mean relative error (in absolute value) of 21% and a systematic bias of -7.7%. The position of the sources in the sky is determined with errors inferior to the size of the pixel. The catalogue of detected sources is complete at fluxes $\geq$ 4.3 Jy. The number of spurious detections is less than a 10% of the true detections. We compared the results with the ones obtained by filtering with a Gaussian filter and a Mexican Hat Wavelet of width equal to the scale of the sources. The adaptive filter outperforms the other filters in all considered cases. We conclude that optimal adaptive filters are well suited to detect and extract sources with a given profile embedded in a background of known statistical properties. In the Planck case, they could be useful to obtain a real-time preliminary catalogue of extragalactic sources, which would have a great scientific interest, e. g. for follow-up observations.