DSARD: DBSCAN-based Solar Active Regions Detection

This is the codebase of the paper Statistical Analyses of Solar Active Regions in SDO/HMI Magnetograms Detected by the Unsupervised Machine Learning Method DSARD .

We propose DSARD (DBSCAN-based Solar Active Regions Detection), an unsupervised machine learning method for the efficient, pixel-level automatic identification of solar active regions (ARs) in magnetograms. It involves an initial thresholding of magnetic field intensities, a two-stage DBSCAN clustering process to identify and refine ARs, and a final integration step that merges regions based on magnetic polarity and proximity while filtering out noise.

How to use the code

• AR_detection_model.py: The basic model for detecting AR using SDO's HMI data. You can utilize it as a demo by changing 'data_item'.
  
• AR_Statistic_get.py: You can run this code to get AR statistics. Settings are define at the beginning of the code.
  
• AR_region_growth.py: We reproduce the previous results of using the region growth algorithm to identify the solar active regions.(Jie Zhang et al.)
  
• AR_csv_data_2010_2023.csv records the information of solar active regions we detected with our algorithm during [2010,2023].

Installation:

conda env create --file environment.yaml
conda activate DSARD
pip install -r requirements.txt
cd DSARD

Run demo:

python -u AR_detection_model.py

Citing

This repository is archived on Zenodo with