Automatic Detection of Sunspots and Solar Filaments from CHASE Data Based on Improved Simulated Annealing Genetic Algorithm

Abstract Sunspots and filaments are critical indicators of solar activity. While the Chinese H α Solar Explorer (CHASE) provides massive full-disk spectral data, existing automated detection algorithms are primarily optimized for other solar observatories, such as the Solar Dynamics Observatory/Helioseismic and Magnetic Imager. The significant discrepancies in instrument responses between different solar observatories render the direct transfer of existing algorithms ineffective, failing to yield robust results when processing CHASE data. To address these challenges, an automatic detection framework based on an Improved Simulated Annealing Genetic Algorithm (ISAGA) is proposed. First, a standardized preprocessing system tailored to CHASE imaging characteristics is established to effectively eliminate limb darkening and transmission artifacts in the Fe I and H α bands. Second, the ISAGA algorithm introduces a physical prior reward mechanism based on the Sigmoid function into the conventional Kapur–Sahoo–Wong (KSW) entropy model, to overcome the shortcomings of handling unimodal histograms, significantly improving the threshold search robustness. Finally, topological postprocessing is used to achieve refined feature extraction. Experiments on the 2022–2023 CHASE dataset demonstrate that ISAGA achieves an intersection over union (IoU) of 98.43% and an F1 score of 99.20% for sunspot detection, outperforming other representative methods. For solar filaments, the framework yields an average IoU of 60.93% and an F1 score of 74.95%. By achieving an effective balance between superior precision and high computational efficiency, ISAGA is suitable for the automated detection of sunspots in massive CHASE data and serves as an effective tool for the rapid generation of large-scale filament datasets.