Answers are grounded in this paper and cite their sources.
Terrestrial laser scanning (TLS) point clouds serve as essential data for 3-D Earth observation, realistic scene reconstruction, and geospatial information modeling. Given their high acquisition costs, they are vulnerable to illegal copying and tampering, making copyright protection increasingly prominent. However, existing zero-watermarking methods generally suffer from drawbacks, such as redundant key-point extraction, susceptibility of features to geometric transformation interference, and insufficient robustness against attacks. To address these problems, a zero-watermarking method for TLS point clouds based on local voxel-center (LVC) features is proposed. First, stable and representative key points are extracted through uniform sampling and dual-threshold screening to reduce redundancy and interference from nonfeature points. Next, a local reference frame combining robust normal vectors and vertical constant vectors is constructed to eliminate the influence of geometric transformations, such as rotation and translation, on feature extraction. Subsequently, LVC features are extracted using a voxel encoding strategy, with voxel centers as the core reference to avoid dependence on single-point coordinates and improve resistance to noise and point-density variations. Finally, a zero watermark with both uniqueness and robustness is generated via principal component analysis dimensionality reduction, K-means clustering normalization, Arnold scrambling, and xor operation. The experimental results show that the proposed method maintains a high copyright authentication accuracy rate under common attack scenarios—such as geometric transformations, cropping, simplification, random point addition, and reordering—with an average normalized correlation coefficient value of 0.9704. It demonstrates excellent robustness and outperforms similar algorithms.
No takes yet. Share an insight, caveat, or question.
Qu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: