A Study on a 3D Model Plagiarism Detection Method Based on Point Cloud Features

This paper proposes a 3D model plagiarism detection method based on point cloud feature analysis to address complex model modification scenarios such as local deformation, geometric distortion, and non-uniform scaling. By converting 3D models into point cloud representations and incorporating an improved Siamese network architecture, the method leverages attention mechanisms and feature fusion strategies to capture geometric similarities and differences between models from both global and local perspectives. Experimental results demonstrate that the proposed method consistently achieves high performance across various transformation scenarios, particularly exhibiting significant advantages in non-uniform scaling cases where traditional digital watermarking and zero-watermarking methods perform poorly. These findings confirm the applicability and robustness of the proposed approach in 3D model copyright protection.