Enhanced precision in automatic identification of Coronal Hole regions in solar images using the proposed Supervised Intensity Thresholding with Distance Transform Clustering and Connected Component Labeling (SITDTCCCL) method on data collected by the Solar Dynamic Observatory between January 2019 and July 2023

Abstract

Coronal Hole (CH) being a crucial feature of solar corona bears immense importance in the field of Astronomy and Solar Physics because it contributes to Geomagnetic storms through the emission of Charged particles into interplanetary space, impacting both space weather and weather of the Earth and also impacts in the lives of Earth and Space. So detection of Coronal Hole regions is a significant task. Many attempts have already been made in this regard. In this work we are proposing a new method, for the automatic detection of CH regions using a deep learning technique, we used Supervised Intensity Thresholding with Distance Transform Clustering and Connected Component Labeling (SITDTCCCL) to find out Regions of Interest (ROI) from solar images of spectrum 193Å193Å of Atmospheric Imaging Assembly (AIA), available at onboard Solar Dynamics Observatory (SDO) and a state-of-the-art deep learning method (three YOLO v 8 models, such as YOLO v8n(nano), YOLO v8m(medium), YOLO v8x(extra large)) which has shown excellent performance in detection of CH regions with the scores of evaluation matrices such as F1 score 95% Precision 97.1%, mAP50 98.1% and True Positive Rate (TPR) 100%.