Abstract: Urban 3D geological modeling in complex tectonic areas faces a trade-off between high structural characterization accuracy and efficient attribute modeling. To address this issue, this study takes the Panlong District of Kunming City as the research area and designs an practically implementable “structure–mesh–attribute” collaborative modeling workflow targeting the practical challenge of “high structural accuracy requirements versus sparse attribute modeling data”. The proposed method integrates Delaunay triangular meshes and truncated rectangular grids. Based on multi-source geological data, the two types of meshes are connected through data transfer constraints and strict topological correspondence to construct an integrated model. The core innovation of this study lies in establishing strict topological correspondence rules (common-point, common-edge, and common-face constraints) between Delaunay triangular meshes and truncated rectangular grids, enabling geometric alignment and attribute continuity at geological interfaces, distinguishing it from conventional weakly constrained mesh fusion methods. The results show that the proposed method can accurately represent complex structures such as faults and folds, improving modeling accuracy by approximately 15% compared to single-mesh methods, while reducing the number of truncated rectangular grid cells by 30% relative to corner-point grids. Leave-one-out cross-validation indicates that the mean absolute error at boreholes is less than 2‰, and the inter-borehole profile data error is less than 2%, with attribute interpolation accuracy meeting the requirements of urban geological engineering applications. This combined mesh generation technique provides reliable technical support for urban geological surveys, underground space planning, and geological hazard assessment in complex tectonic areas, and offers a standardized grid data carrier for the subsequent introduction of deep learning and semi-supervised learning.