Barkai, Jordan (2022) A comparative study of source finding techniques in H i emission cubes using SoFiA, MTObjects and supervised deep learning. Master's Thesis / Essay, Astronomy.
|
Text
Masters_Thesis_to_submit.pdf Download (5MB) | Preview |
|
|
Text
toestemming.pdf Download (117kB) | Preview |
Abstract
Astronomical surveys map the skies without a specific target, resulting in images containing many astronomical objects. As the technology used to create these surveys improves with projects like the SKA, an unprecedented amount of data will become available. Hence the need for fast and accurate techniques to detect and locate sources in astronomical survey data. The challenge lies in the lack of clarity in the boundaries of sources, with many having intensities very close to the noise, especially in the case of radio data. This project therefore aims to find the best source finding pipeline for 3D neutral hydrogen cubes from the Westerbork Synthesis Radio Telescope (WSRT). This was achieved by first testing two traditional source finding methods, the well established HI source finding tool SoFiA and one of the latest best performing optical source finding tools, MTObjects. A new supervised deep learning approach was also tested, in which a 3D convolutional neural network architecture, known as V-Net, originally designed for medical imaging was used. These three source finding methods were also further improved by adding classical machine learning classifiers as a post-processing step to remove any false positive detections.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Verheijen, M.A.W. and Talavera Martinez, E. and Wilkinson, M.H.F. |
| Degree programme: | Astronomy |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 26 Jan 2022 10:09 |
| Last Modified: | 15 Feb 2024 14:36 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/26502 |
Actions (login required)
 |
View Item |