| Name: | Pranava Seth |
| Affiliation: | Thapar Institute of Engineering and Technology (TIET) |
| Conference ID : | ASI2024_323 |
| Title : | An Artificial Intelligence (A.I.) based chromospheric feature extractor and classifier for SUIT. |
| Authors : | Pranava Seth, Megha Anand, Vishal Upendran, Soumya Roy, Durgesh Tripathi |
| Authors Affiliation: | Pranava Seth Affiliation (Thapar Institute of Engineering & Technology (TIET), Patiala-147004, India/ Inter-University Centre for Astronomy & Astrophysics (IUCAA), Pune - 411007, India ).
Megha Anand, Soumya Roy, Durgesh Tripathi Affiliation (Inter-University Centre for Astronomy & Astrophysics (IUCAA), Pune - 411007, India).
Vishal Upendran Affiliation (Bay Area Environmental Research Institute (BAERI), California -94035, USA /Lockheed Martin Solar and Astrophysics Laboratory (LMSAL), California - 94304, USA).
|
| Mode of Presentation: | Poster |
| Abstract Category : | Sun, Solar System, Exoplanets, and Astrobiology |
| Abstract : | The Solar Ultraviolet Imaging Telescope (SUIT) onboard Aditya-L1 will observe the Sun in the wavelength range of 200-400 nm covering the Solar photosphere and chromosphere features. SUIT will provide routine partial and full disk images of the Sun in 11 different science filters. It will observe solar features like plages, sunspots, filaments, etc. An accurate understanding of the plasma and thermodynamic properties of these regions requires the development of automatic feature detection methods. The primary objective of this work is to detect and classify the Solar chromospheric features to be observed from SUIT’s Mg II filter. Specifically, we target plage regions, sunspots, filaments, and off-limb structures like prominences. The system under consideration employs a hybrid methodology that integrates computer vision, supervised and unsupervised machine learning. For this purpose, we use Interface Region Imaging Spectrometer (IRIS) full-disk mosaic images in Mg II h and k lines. Multiple pre-processing strategies are employed to eliminate image artifacts and noises from the IRIS mosaic data. A custom dataset is created from the cleaned data incorporating SUIT PSF, pixel resolution (0.7”), and various augmentations. The proposed system utilizes a two-step detection approach. The first module identifies the regions of interest from the cleaned dataset using a YOLO v5 neural network model. The subsequent module integrates the techniques of supervised classification and clustering to analyze multiple numerical features, which act as an initial event confirmation system. This pipeline when deployed will provide a real-time display of the detection results, as well as the time of recording, pixel coordinates, and Helioprojective Coordinates of the classified features to be observed from the SUIT full-disk Mg II images.
|
