Independent Verification for Traditional Automated Sunspot Recognition Across Multiple Solar Observations

Abstract

Automated sunspot recognition is critical for understanding solar activity and its space weather impacts. However, independent verification of recognition results remains a challenge for traditional methods. In this study, an independent verification strategy for traditional automated sunspot recognition using an improved LeNet-5 network was proposed and assessed across multiple solar observations. First, a hybrid recognition method combining threshold segmentation and mathematical morphology was applied to SOHO/MDI full-disk white-light intensity continuum images spanning the entire Solar Cycle 23. To independently verify the recognition results, an improved LeNet-5 network was employed as an objective verifier, trained on a balanced dataset of these automatically recognized sunspot and background samples. The classifier achieved accuracies of 99.67% and 99.76% on validation and testing sets, respectively. Subsequently, sunspot features, including number and area, were extracted and compared with the solar region summary records. The Spearman and Pearson correlation coefficients for sunspot number were 0.8710 and 0.8281, respectively, and those for sunspot area reached 0.9386 and 0.9368. These results collectively provide a comprehensive assessment of the sunspot recognition performance through both deep-learning classification and statistical comparison. Furthermore, to evaluate robustness and applicability, the same procedure was applied to the reduced-resolution SDO/HMI images from the overlapping period with SOHO/MDI during Solar Cycle 24. Classification accuracies remained above 99%, and feature consistency between instruments was high. Although correlation coefficients during this low-activity overlap period were lower than those of full Solar Cycle 23, the results from HMI closely tracked those from MDI. These findings confirm that the proposed independent verification approach enhances the reliability of traditional recognition algorithms and maintains consistency across multiple solar observations.