Abstract: 【Objective】3D geological model data has the characteristics of multi-scale, multi-attribute, and uneven spatial distribution, with complex geological body morphology and large data volume. The commonly used R-tree and Hilbert R-tree indexes have problems such as high leaf node overlap, poor spatial clustering, and insufficient query locality when processing such data, making it difficult to meet the requirements of efficient random access.【Methods】Therefore, a clustering Hilbert R-tree based indexing method is proposed. This method constructs a spatial indexing system for geological body objects by combining the partitioning strategy of geological spatial objects with a two-stage clustering strategy coupled with the spatial filling characteristics of Hilbert curves. On the basis of traditional Hilbert R-tree, K-means clustering algorithm is introduced to perform minimum bounding rectangle (MBR) geometric clustering on leaf nodes, and then perform Hilbert code value semantic clustering on non-leaf nodes, enhancing the spatial compactness and clustering balance of Hilbert R-tree leaf nodes, reducing the redundant area of MBR and the spatial overlap between adjacent nodes in case of uneven data distribution, thus achieving efficient range query.【Results】Using a large-scale multi-scale 3D geological model in Guizhou for testing, when processing large-scale datasets, its index space occupation is reduced by 17%, query response time is increased by 35%, and query efficiency is significantly improved.【Conclusion】The results show that compared with traditional Hilbert R-tree, the clustering Hilbert R-tree indexing method has significant improvements in index spatial proportion and query response time.