基于特征点提取的三维点云模型水印算法

为了有效提高三维水印的透明性、抗噪能力和水印提取准确度，针对三维网格中不固定点云数据，提出一种基于局部特征点提取的三维点云模型水印算法。根据协方差分析提取出三维模型初始特征点，以初始特征点为核心，在它K近邻邻域中，构建不跨越区域最小三角形为嵌入单元的底面，将剩余顶点按照升序排列，寻找合适的嵌入顶点，构建局部嵌入单元，通过改变嵌入顶点信息来嵌入水印。其中通过顶点在平面投影产生的夹角确定水印索引值，实现盲水印。算法通过保留特征点信息，改变非特征点嵌入水印信息能有效提高透明性和抗噪能力，通过限制嵌入单元区域提高水印提取准确率，同时实现了盲水印检测。

3D point cloud model watermarking algorithm based on feature points extraction

To effectively improve the transparency and noise immunity of three-dimensional watermark and accuracy of watermark extraction, for the unfixed point cloud data of three-dimensional grid, this paper proposed an three-dimensional point cloud watermarking algorithm based on feature points extraction. According to analysis of covariance, initial feature points were extracted from point cloud model. Every initial feature point was the core point of K-nearest neighbor local neighborhood to build no cross-regional and the smallest triangle, which was the embedding unit bottom, surface, and then the remaining vertices were arranged in ascending order. One suitable embedding vertex was searched as an embedding vertex. By changing the vertex information of local embedding unit, watermark was embedded. Blind watermark was achieved by the index values which was generated by the angle of the vertex's projection. Keeping feature point information, the proposed algorithm effectively improved transparency and noise immunity. By changing the non-characteristic point to embed watermark information and limiting the area of embedding unit, the proposed algorithm improves the watermark accuracy and achieves the blind watermark detection.