Abstract: The geometric morphology and distribution characteristics of mineral particles are of great significance for understanding the genesis of rocks and the evolution process of deposits. Traditional miner⁃ al labeling methods rely on manual operation，which is time-consuming and susceptible to subjective er⁃ rors，and it is difficult to meet the needs of batch，efficient and accurate analysis. In order to solve this problem，this paper takes the granite of Maoping deposit in Jiangxi Province as the research object，and proposes a mineral morphology research framework combining image large model and fractal analysis. The Segment Anything Model（SAM）image large model is used to segment the mineral microscopic image effi⁃ciently and intelligently，and the single fractal and multifractal analysis of the segmented mineral particles is carried out. The morphological characteristics of minerals are quantitatively characterized by box dimen⁃ sion（Db ），perimeter-area dimension（DPA），information dimension（D1 ），correlation dimension（D2 ） and multifractal parameters. The results show that compared with manual labeling，SAM intelligent seg⁃ mentation significantly improves the labeling efficiency while achieving high accuracy. Through single frac⁃ tal analysis，the morphology and distribution characteristics of granite mineral particles are revealed from multiple angles. The Db of quartz is significantly lower than that of other minerals，indicating that its parti⁃ cle distribution is more uniform and its morphology is more regular. The DPA of mica is significantly higher than that of other minerals，reflecting that its particle boundary is complex and irregular. The multifractal analysis further reveals the inhomogeneity and complexity of mineral distribution. The average values of D1 and D2 of quartz are significantly higher than those of other minerals，indicating that its spatial distribution has strong self-similarity. The multifractal spectrum width ∆α of plagioclase is the largest，indicating that the heterogeneity of its probability distribution is higher than that of other minerals. In addition，Δf（α） characterizes the dominant characteristics of mineral particles at different scales in the multifractal struc⁃ ture. The results show that the Δ（f α）values of most minerals are positive，indicating that medium-coarse particles dominate the fractal structure. Finally，in the DPA-Db coordinate diagram，each mineral showed a significant distribution pattern according to the morphological characteristics，and achieved effective classi⁃ fication and identification. The framework based on image large model and fractal analysis shows certain ac⁃ curacy and efficiency in characterizing the morphology and distribution pattern of mineral particles，which provides important support for the study of mineral morphology and genetic process.

Key words: image large models, fractal analysis, morphology of mineral particles, distribution pat? tern of mineral, mineral classification