为提高三维模型版权信息的安全性和鲁棒性, 提出了一种基于尺度不变特征转换的三维点云模型水印算法。首先对三维点云模型进行主成分分析, 然后将模型中的点到模型中心的距离r进行一定的排序, 形成一个m×n的矩阵。对该矩阵进行尺度不变特征提取, 找到其稳健的特征点。在选取合适的特征点后对其邻域进行Contourlet变换, 将水印嵌入其低频部分, 然后进行Contourlet逆变换, 即可生成带水印的三维点云模型。实验表明, 该算法对几何攻击以及噪声、裁切、简化等攻击具有较强的鲁棒性。该算法可用于实际三维模型的版权认证。

A novel algorithm for 3D point cloud model based on scale-invariant feature transform is presented to improve the security and robustness of copyright.Firstly, principal component analysis is used for 3D point cloud model, and then the distances from the points to the center in the model are in specific sorting to form a matrix. The scale invariant feature points of the matrix can be extracted by scale-invariant feature transform and the steady feature points can be found in the feature points.Contourlet transform is performed on neighborhood of the steady feature points, and watermark is embedded in the low frequency of the neighborhood, and then Contourlet inverse transform is performed to get a 3D point cloud model with watermark.Experiments show that the algorithm has strong robustness to geometric attacks and attacks such as cutting, noise, simplify. The proposed algorithm can be used for the practical 3D model of copyright authentication.