Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 2: AKPIK II: Deep Learning

AKPIK 2.7: Vortrag

Mittwoch, 17. März 2021, 17:30–17:45, AKPIKa

Kinematic Analysis of Radio Jets with Deep Learning — •Paul-Simon Blomenkamp and Kevin Schmidt — TU Dortmund, Dortmund, Deutschland

Active galactic nuclei (AGN) are some of the most intensely studied objects in the night sky. Some of these AGNs can accelerate matter in their core to relativistic speeds. These jets are prominent sources in radio astronomy. Analysing the kinematic properties of radio jets can give insight on many physical properties of the host galaxy. Previously this analysis was mostly done by manually tracking Gaussian components of the jets, which by its nature involves some degree of arbitrariness.

This work aims at the automated detection of Gaussian components in radio jets with Deep Learning. This is expected to accelerate and improve on the manual methods. Ideally, the model will be able to independently identify the components and their position in order to perform a kinematic analysis. To achieve these aims a Deep Learning model using Convolutional Neural Networks is to be developed. The current results have been achieved by using object detection models. The used dataset is composed of simulated Gaussian jet components in 640× 640 images and is labelled. At the current state, the model is able to confidently identify and locate all the Gaussian components within the image.