Automatic Detection of Sunspots on full-disk continuum images using the MiniMax Optimization and Feature Extraction

Abstract

Novel contributions engaging white and black-box models for Sunspot detection neglect optimization which is highly essential in solar activity monitoring. This demanded the quest for automated mathematical programming methods, warranting accurate forecasting of solar activity. Accordingly, in this paper, an energy-optimized adaptive curve fitting sunspot detection model coined Sunspot Attribute Extraction via Energy-tuned Minimax Optimization (SAEEMO) is introduced for characterizing diverse solar constituents followed by feature extraction from full disk images. SAEEMO initially considers the overall image energy as the sum of energies concerned with the solar disk to distinguish the solar variations that are intensity enhanced, facilitating segmentation. Later, the Sunspots are further disintegrated into umbra and penumbra, characterized as edge sensitivity and regularization terms, and adaptively thresholded using the novel energy-based Minimax optimizer. The adopted energy function packs the intensity variations within the Minimax bounds to render a globally optimal solution in an iterative manner using the line search operation. SAEEMO’s Receiver Operating Characteristics (ROC) investigation on Helioseismic and Magnetic Imager (HMI) continuum images obtained from Solar Dynamics Observatory (SDO) reveals its preservative and distinctive nature in detecting and characterizing diverse solar features. Also, relative investigations of the extracted solar features with the catalogs of Debrecen Photoheliographic Data (DPD), Heliophysics Feature Catalog (HFC), Sunspot Index and Long-term Solar Observations (SILSO), and Space Weather Prediction Center (SWPC) demonstrate SAEEMO’s consistency.