Prospects on high-energy source searches based on pattern recognition: Object detection in the H.E.S.S. Galactic Plane Survey and catalogue cross-matches

Abstract The H.E.S.S. Galactic Plane Survey [HGPS, 1] represents one of the most sensitive surveys of the Galactic Plane at very high energies (VHE, 0.1 Our goal is to build in a short amount of computational time a list of potentially valuable objects without prior case-specific morphological assumptions. We aim to classify and rank the detected objects in order to identify only the most promising source candidates for further multi-wavelength-association searches, dedicated analyses, or deeper observations. In the approach proposed, we extract sparse and pertinent structural information from the significance maps using a edge detection operator. We then apply a Hough circle transform and detect a collection of objects as local maxima in the Hough space. On the basis of morphological parameters we can characterize different object classes. Classification can be used to identify valuable source candidates sharing the characteristics of well-known sources. We show that using these pattern recognition techniques we can detect objects with partial circular symmetry irrespective of a morphological template (e.g. point-like, Gaussian-like, or shell-like). All the shell-type supernova remnants (SNRs) catalogued in the HGPS (from dedicated analyses) are associated with at least one detected object. Catalogue cross-matches indicate that several detected objects not catalogued in the HGPS are spatially coincident with multi-wavelength counterparts. This paper can be seen as a prospective study for the search of VHE γ-ray sources based on Hough transform and morphological classification. The algorithm have been tested on bootstrap simulations and applied to significance maps of the H.E.S.S. Galactic survey. Further investigation on the most promising candidates will be conducted in dedicated follow-up analyses.