Study on correlation between lunar surface annihilation radiation fluence rate and lithology based on Fuzzy C-means Gaussian Mixture Method

To solve the problem of incomplete mapping between lunar surface lithology and annihilation radiation flux, an adaptive mixed clustering method called Fuzzy C-Means Gaussian Mixture Clustering (FCGMM) is proposed. A mathematical model for the inversion of typical rock types using lunar surface annihilation radiation flux rate was constructed through adaptive initialization processing of non-equilibrium data training sets using hard labels. The misclassification rates of six rock types were calculated using a layered K-fold cross validation method. The average misclassification rates of ferrous plagioclase, low titanium basalt, ferrous norite, and high titanium basalt were 1.34%, 7.63%, 9.67%, and 6.68%, respectively. The misclassification rates of magnesian plagioclase and Kripkite were slightly higher, at 10.27% and 10.82%, respectively. The FCGMM method not only significantly improves accuracy compared to FCM and GMM, but also achieves high-precision inversion of the monthly spatial distribution of six types of rocks under training conditions with small data samples (20%). The inversion results of the entire month are highly consistent with the comprehensive measurement results of spectra, radar, neutrons, etc., with a relative error of -0.95% to 3.97%, confirming the scientific validity of using "annihilation radiation" (a single parameter) to invert lithology results. Based on this discovery, there are magnesian plagioclase rocks present in the Moscow Sea and Bayi Crater, while basalt may be distributed in the Letrona Crater. The research results further demonstrate the correlation between lunar surface annihilation radiation flux and lithology, achieving deep mining and scientific application of lunar surface annihilation radiation flux data.