<p>In the era of rapid advancements in smart cities and digital technology, the application of oblique photography 3D models has become increasingly widespread, and data security issues have garnered significant attention. However, most existing watermarking algorithms lack strong robustness against scaling, cropping, and composite attacks. To address this problem, this paper proposes an oblique photography 3D model watermarking algorithm based on multi-level curvature and weighted distance grouping. First, the curvature values of vertices in different neighborhood levels are calculated using principal component analysis (PCA), and vertices with curvature values exceeding a predefined threshold are selected as feature points. Next, vertices are grouped according to the principle of minimizing the weighted comprehensive distance between feature points and vertices. Finally, the height difference between each vertex and its corresponding feature point within a group is normalized to serve as the carrier for embedding watermark information, and the Quantization Index Modulation (QIM) method is employed for watermark embedding. Experimental results show that the algorithm achieves high imperceptibility and demonstrates strong robustness against common attacks such as geometric transformations, vertex reordering, and composite attacks. Even under cropping attacks of up to 80%, the complete watermark information can still be successfully retrieved.</p>