New Method Enhances Detection of Solar Active Regions

Recent advancements in solar astrophysics have been made with the introduction of a new method for detecting Solar Active Regions (ARs). This technique utilizes images from NASA's Solar Dynamics Observatory (SDO) and employs a combination of 2D circular kernel time series transformation, statistical measures, entropy measures, and machine learning algorithms. The method transforms the circular area around pixels in SDO AIA images into one-dimensional time series (1-DTS).

The study, titled "Solar Active Regions Detection Via 2D Circular Kernel Time Series Transformation, Entropy and Machine Learning Approach," outlines how statistical measures such as Median Value and 95th Percentile, along with entropy measures like Distribution Entropy and Fuzzy Entropy, are used for feature selection. The machine learning model classifies these time series into three categories: no Active Region, non-flaring Regions outside active regions, and flaring Active Regions.

The results indicate a classification accuracy of 90.0% for entropy measures and 91.4% for statistical measures, with Fuzzy Entropy achieving the highest accuracy of 89.5%. This suggests that the proposed method is effective for detecting ARs in SDO AIA images, which could enhance our understanding of solar activity and its implications for space weather. The findings are significant as they may improve predictive capabilities regarding solar flares and other solar phenomena that can impact Earth.

The full paper can be accessed at arXiv:2306.08270.