2022MNRAS.510.4031L

We demonstrate two new approaches that have been developed to aid the production of future hard X-ray catalogues, and specifically to reduce the reliance on human intervention during the detection of faint excesses in maps that also contain systematic noise. A convolutional neural network has been trained on data from the INTEGRAL/ISGRI telescope to create a source detection tool that is more sensitive than previous methods, whilst taking less time to apply to the data and reducing the human subjectivity involved in the process. This new tool also enables searches on smaller observation time-scales than was previously possible. We show that a method based on Bayesian reasoning is better able to combine the detections from multiple observations than previous methods. When applied to data from the first 1000 INTEGRAL revolutions these improved techniques detect 25 sources (about 5 per cent of the total sources) which were previously undetected in the stacked images used to derive the published catalogue made using the same data set.