A comparative study of source-finding techniques in H I emission line cubes using SoFiA, MTObjects, and supervised deep learning

Abstract

Context. The 21 cm spectral line emission of atomic neutral hydrogen (H I) is one of the primary wavelengths observed in radio astronomy. However, the signal is intrinsically faint and the H I content of galaxies depends on the cosmic environment, requiring large survey volumes and survey depth to investigate the H I Universe. As the amount of data coming from these surveys continues to increase with technological improvements, so does the need for automatic techniques for identifying and characterising H I sources while considering the tradeoff between completeness and purity. Aims. This study aimed to find the optimal pipeline for finding and masking the most sources with the best mask quality and the fewest artefacts in 3D neutral hydrogen cubes. Various existing methods were explored, including the traditional statistical approaches and machine learning techniques, in an attempt to create a pipeline to optimally identify and mask the sources in 3D neutral hydrogen (H I) 21 cm spectral line data cubes. Methods. Two traditional source-finding methods were tested first: the well-established H I source-finding software SoFiA and one of the most recent, best performing optical source-finding pieces of software, MTObjects. A new supervised deep learning approach was also tested, in which a 3D convolutional neural network architecture, known as V-Net, which was originally designed for medical imaging, was used. These three source-finding methods were further improved by adding a classical machine learning classifier as a post-processing step to remove false positive detections. The pipelines were tested on H I data cubes from the Westerbork Synthesis Radio Telescope with additional inserted mock galaxies. Results. Following what has been learned from work in other fields, such as medical imaging, it was expected that the best pipeline would involve the V-Net network combined with a random forest classifier. This, however, was not the case: SoFiA combined with a random forest classifier provided the best results, with the V-Net–random forest combination a close second. We suspect this is due to the fact that there are many more mock sources in the training set than real sources. There is, therefore, room to improve the quality of the V-Net network with better-labelled data such that it can potentially outperform SoFiA.